Phase Change in Mechanics

Lecture Notes of the Unione Matematica Italiana

For further volumes: http://www.springer.com/series/7172

13

Editorial Board

Franco Brezzi (Editor in Chief) IMATI-CNR Via Ferrata 5a 27100 Pavia, Italy e-mail: [email protected] John M. Ball Mathematical Institute 24-29 St Giles’ Oxford OX1 3LB United Kingdom e-mail: [email protected] Alberto Bressan Department of Mathematics Penn State University University Park State College PA. 16802, USA e-mail: [email protected] Fabrizio Catanese Mathematisches Institut Universitatstraße 30 95447 Bayreuth, Germany e-mail: [email protected] Corrado De Concini Dipartimento di Matematica Universit´a di Roma “La Sapienza” Piazzale Aldo Moro 2 00185 Roma, Italy e-mail: [email protected] Persi Diaconis Department of Statistics Stanford University 450 Serra Mall Stanford, CA 94305-4065, USA e-mail: [email protected], [email protected]

Nicola Fusco Dipartimento di Matematica e Applicazioni Universit´a di Napoli “Federico II”, via Cintia Complesso Universitario di Monte S. Angelo 80126 Napoli, Italy e-mail: [email protected] Carlos E. Kenig Department of Mathematics University of Chicago 5734 University Avenue Chicago, IL 60637-1514 USA e-mail: [email protected] Fulvio Ricci Scuola Normale Superiore di Pisa Piazza dei Cavalieri 7 56126 Pisa, Italy e-mail: [email protected] Gerard Van der Geer Korteweg-de Vries Instituut Universiteit van Amsterdam Plantage Muidergracht 24 1018 TV Amsterdam, The Netherlands e-mail: [email protected] C´edric Villani Institut Henri Poincar´e 11 rue Pierre et Marie Curie 75230 Paris Cedex 05 France e-mail: [email protected]

The Editorial Policy can be found at the back of the volume.

Michel Fr´emond

Phase Change in Mechanics

123

Michel Fr´emond Universit`a di Roma “Tor Vergata” Dipartimento di Ingegneria Civile Via Politecnico 1 00133 Roma Italy [email protected]

ISSN 1862-9113 ISBN 978-3-642-24608-1 e-ISBN 978-3-642-24609-8 DOI 10.1007/978-3-642-24609-8 Springer Heidelberg Dordrecht London New York Library of Congress Control Number: 2011945415 Mathematics Subject Classification (2010): 74N25; 74R20; 74M15; 74M20; 80A17; 86A10; 70F35; 76A02; 35Q35; 35Q74 c Springer-Verlag Berlin Heidelberg 2012  This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

A model is a theory which predicts the evolution of a structure. The basic elements of a mechanical predictive theory are described. They involve the choice of the state quantities which characterize the investigated physical phenomenon, the basic equations of mechanics: the equations of motion, the laws of thermodynamics and the constitutive laws. The theory is applied to phase change. Phase change involves microscopic motions which have macroscopic effects. These motions are taken into account in the macroscopic predictive theories. Numerous examples which are used or may be used in engineering are given for different phase change and damage problems. Moreover, when phase change occurs, temperatures may be discontinuous, for instance when warm rain falls on frozen soil. This situation is also investigated. Some of these problems have been investigated in the framework of the Laboratorio Lagrange with Italian, Tunisian and French scientists, Francesco Ascione, Elena Bonetti, Anna Maria Caucci, Eric Dimnet, Christian Duquennoi, Francesco Freddi, Rym Lassoued, Boumediene Nedjar, Francesca Nerilli and Elisabetta Rocca. All of them are warmly thanked. These lecture notes are based on a course given at the XXX Scuola Estiva di Fisica Matematica at Ravello in September 2005 and on lectures given at the Universit`a di Roma “Tor Vergata”. The author would like to thank the Scuola Estiva di Fisica Matematica, the Universit`a di Roma “Tor Vergata”, the Unione Matematica Italiana, and Professors Mauro Fabrizio, Franco Maceri and Franco Brezzi for the opportunity to give these lectures and to publish them. Roma, Italy

Michel Fr´emond

v

•

Contents

1

Introduction .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

1

2

The State Quantities and the Quantities Describing the Evolution . . . 2.1 The State Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 2.2 The Quantities Which Describe the Evolution and the Thermal Heterogeneity . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

3 3

3

4

The Basic Laws of Mechanics . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.1 The Principle of Virtual Power .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.2 The Laws of Thermodynamics .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.2.1 The Energy Balance . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.2.2 The Second Law of Thermodynamics . . . . . . . . . . . . . . . . . . 3.2.3 The Free Energy . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.2.4 Equivalent Laws of Thermodynamics . . . . . . . . . . . . . . . . . . 3.2.5 The Entropy Balance Equivalent to the Energy Balance . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.2.6 An Equivalent Second Law of Thermodynamics . . . . . . 3.3 The Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.3.1 An Extra State Quantity . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.4 Internal Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.4.1 Constitutive Laws in Case There Are Internal Constraints on the State Quantities . . . . . . . 3.4.2 Some Notations . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.5 The Basic Laws of Mechanics on a Discontinuity Surface . . . . . . . 3.6 The Basic Laws of Mechanics in a Discontinuity Volume .. . . . . . . 3.6.1 The Principle of Virtual Power and the Equations of Motion . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3.6.2 The Laws of Thermodynamics .. . . . .. . . . . . . . . . . . . . . . . . . . Solid–Liquid Phase Change . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.1 The Stefan Problem .. . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.1.1 The Equations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

4 5 5 8 8 9 9 9 10 11 11 13 13 15 19 19 24 24 25 31 31 33 vii

viii

Contents

4.1.2

4.2 4.3

4.4 4.5

4.6

4.7 4.8 4.9

A Way to Solve the Equations Where the Free Boundary Is a By-Product: The Freezing Index . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Unfrozen Water in the Frozen Zone.. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.3.1 The Equations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.3.2 An Example . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Phase Change with a Freezing Fringe.. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.4.1 The Equations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Phase Change with Voids and Bubbles . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.5.1 The Equations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.5.2 The Case Where There Is Only Liquid: The Cavitation Phenomenon . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.5.3 The Case There Is Only Solid: Soil Mechanics . . . . . . . . 4.5.4 Numerical Examples.. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.5.5 Example 1: The Cavitation Phenomenon .. . . . . . . . . . . . . . 4.5.6 Example 2: The Non Dissipative Spatial Interaction Is Important.. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.5.7 Example 3: No Spatial Interaction . .. . . . . . . . . . . . . . . . . . . . 4.5.8 Example 4: Dissipative Phase Change .. . . . . . . . . . . . . . . . . 4.5.9 Example 5: Non Dissipative Phase Change . . . . . . . . . . . . 4.5.10 Example 6: Effect of the Pressure .. .. . . . . . . . . . . . . . . . . . . . Solid–Liquid Phase Change with Different Densities . . . . . . . . . . . . . 4.6.1 The Mass Balance . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.6.2 The Equations of Motion . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.6.3 The Entropy Balance . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.6.4 The Constitutive Laws . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.6.5 Approximation of the Equation for the Macroscopic Motions .. . . . . . .. . . . . . . . . . . . . . . . . . . . 4.6.6 The Equations for the Phase Change.. . . . . . . . . . . . . . . . . . . 4.6.7 An Example: Liquid Water in an Impermeable Container.. . . . . . .. . . . . . . . . . . . . . . . . . . . 4.6.8 An Other Example: Ice in an Impermeable Container . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.6.9 An Other Example: Freezing of a Water Emulsion in an Impermeable Container . . . . . . . . . . . . . . . . 4.6.10 The Non Homogeneous Evolution ... . . . . . . . . . . . . . . . . . . . 4.6.11 The Motion of Glaciers . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Dissipative and Irreversible Phase Change . . . .. . . . . . . . . . . . . . . . . . . . 4.7.1 The Equations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . The Phase Volume Fraction Depends on the Temperature .. . . . . . . 4.8.1 The Equations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Phase Change with Thermal Memory . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 4.9.1 The Equations .. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

34 36 36 38 39 40 40 41 45 45 46 47 48 48 48 50 50 51 53 53 53 54 54 56 56 58 59 61 61 62 62 63 63 64 65 65

Contents

5

ix

Shape Memory Alloys .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.2 The State Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.3 The Equations of Motion . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.3.1 The Equations for the Microscopic Motions When Voids and Interpenetration Are Possible . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible .. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.4.1 The Entropy Balance . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.4.2 The Free Energy . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.4.3 The Constitutive Laws . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.4.4 An Example of Non Dissipative Evolution .. . . . . . . . . . . . 5.4.5 Latent Heat of Austenite–Martensite Phase Change . . . 5.4.6 The Dissipative Constitutive Laws . .. . . . . . . . . . . . . . . . . . . . 5.4.7 Evolution of a Shape Memory Alloy . . . . . . . . . . . . . . . . . . . 5.5 Education of Shape Memory Alloys . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.5.1 The Two-Shape Memory Effect .. . . .. . . . . . . . . . . . . . . . . . . . 5.5.2 A Predictive Theory of the Education of Shape Memory Alloys . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.6 Shape Memory Alloys with the Possibility of Voids . . . . . . . . . . . . . . 5.6.1 The State Quantities . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.6.2 The Mass Balance . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.6.3 The Equations of Motion . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.6.4 The Free Energy . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.6.5 The Pseudopotential of Dissipation .. . . . . . . . . . . . . . . . . . . . 5.6.6 The Constitutive Laws . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.6.7 The Entropy Balance . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5.6.8 The Set of Partial Differential Equations . . . . . . . . . . . . . . . 5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

67 67 67 68

84 91 92 92 93 93 95 96 98 99 99

6

Damage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 6.1 The Equations of Motion . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 6.2 Free Energy and Pseudopotential of Dissipation . . . . . . . . . . . . . . . . . . 6.3 The Equations .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 6.4 The Macroscopic Motions Become Microscopic . . . . . . . . . . . . . . . . . 6.5 An Example: The Wedge Splitting Test . . . . . . .. . . . . . . . . . . . . . . . . . . .

101 102 102 103 104 107

7

Contact with Adhesion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7.1 The Equations of Motion . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7.2 The Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7.3 The Equations .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 7.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

109 110 111 112 113

68 69 70 70 74 76 77 78 79 80 82

x

8

9

Contents

Damage of Solids Glued on One Another: Coupling of Volume and Surface Damages .. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.1 State Quantities and Quantities Describing the Evolution . . . . . . . . 8.2 Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.2.1 Virtual Power of the Interior Forces . . . . . . . . . . . . . . . . . . . . 8.2.2 Virtual Power of the Exterior Forces .. . . . . . . . . . . . . . . . . . . 8.2.3 Virtual Power of the Acceleration Forces.. . . . . . . . . . . . . . 8.2.4 The Principle of Virtual Power .. . . . .. . . . . . . . . . . . . . . . . . . . 8.2.5 The Equations of Motion . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.3 The Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.4 The Equations .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.4.1 On the Contact Surface . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.4.2 In the Domains .. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.5.1 Four Points Bending . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.5.2 Pull Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 8.5.3 Fibre-Reinforced Polymers-Concrete Delamination . . .

115 115 116 116 118 118 118 119 119 122 122 123 124 124 126 127

Phase Change with Discontinuity of Temperature: Warm Water in Contact with Cold Ice . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9.1 Warm Water in Contact with Cold Ice . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9.2 Mixture of Ice and Water in Contact with Cold Ice . . . . . . . . . . . . . . . 9.3 An Example: Rain Falling on a Frozen Ground .. . . . . . . . . . . . . . . . . . 9.3.1 A Dissipative Behaviour .. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

131 131 133 134 136

10 Phase Change and Collisions . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.1 Collisions of Two Balls: The Mechanical Theory .. . . . . . . . . . . . . . . . 10.1.1 The Velocities of Deformation . . . . . .. . . . . . . . . . . . . . . . . . . . 10.1.2 The Principle of Virtual Work . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.1.3 The Virtual Works . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.1.4 The Equations of Motion . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.1.5 An Example of Collisions: The Equations for U + . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.1.6 An Other Example: The Impenetrability Condition .. . . 10.2 Collisions of Two Balls: The Thermal Phenomenon .. . . . . . . . . . . . . 10.2.1 First Law of Thermodynamics for a Point .. . . . . . . . . . . . . 10.2.2 Second Law of Thermodynamics for a Point .. . . . . . . . . . 10.2.3 A Useful Inequality for a Point. . . . . .. . . . . . . . . . . . . . . . . . . . 10.2.4 The First Law for the System . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.2.5 The Second Law for the System . . . .. . . . . . . . . . . . . . . . . . . . 10.2.6 A Useful Inequality for the System .. . . . . . . . . . . . . . . . . . . . 10.2.7 The Constitutive Laws . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.2.8 An Example of Thermal Effects Due to Collisions . . . . 10.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

141 141 142 143 144 145 147 150 153 154 154 155 155 156 157 158 159 162

Contents

10.4

xi

The Thermomechanical Theory Involving Phase Change . . . . . . . . 10.4.1 The Equations of Motion . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.4.2 The Laws of Thermodynamics for Each Ball . . . . . . . . . . 10.4.3 The Laws of Thermodynamics for the System . . . . . . . . . 10.4.4 The Constitutive Laws . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Examples of Thermal Effects with Phase Changes.. . . . . . . . . . . . . . . 10.5.1 The Thermal Equations in the Isentropic Case . . . . . . . . . 10.5.2 The Thermal Equations in the Adiabatic Case . . . . . . . . . 10.5.3 The Adiabatic Situation with the Small Perturbation Assumption . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.5.4 Identical Balls at the Same Temperature Before the Collision . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.5.5 Identical Balls at Different Temperatures Before the Collision: Collision of Two Pieces of Ice . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.5.6 Identical Balls at Different Temperatures Before the Collision: Collision of a Droplet of Water with a Piece of Ice . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 10.5.7 A Droplet of Rain Falls on a Frozen Ground: Does It Give Black Ice? . . . . . . . . . . . . . . . . . . . . . . .

163 163 165 168 172 174 176 177

11 Collisions of Deformable Bodies and Phase Change .. . . . . . . . . . . . . . . . . . . 11.1 The Principle of Virtual Work . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11.2 The First Law of Thermodynamics . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11.3 The Second Law of Thermodynamics . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11.4 The Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11.5 Evolution in a Collision . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11.5.1 The Mechanical Evolution When Decoupled from the Thermal Evolution . . . . . . . . . . . . . . . . 11.5.2 An Example: Collision of a Bar with a Rigid Support .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 11.5.3 Thermal Evolution When the Mechanical Equations Are Decoupled from the Thermal Equations.. . . . . . . .. . . . . . . . . . . . . . . . . . . . 11.5.4 The Temperature Variation in a Collision . . . . . . . . . . . . . .

193 193 195 202 204 207

211 214

12 Phase Change Depending on a State Quantity: Liquid–Vapor Phase Change .. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 12.1 The Temperature Is Lower than Critical Temperature Tc . . . . . . . . . 12.1.1 Partial Phase Change . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 12.2 The Temperature Is Larger than Critical Temperature Tc . . . . . . . . .

217 218 221 221

10.5

177 179

181

189 190

209 210

13 Clouds: Mixture of Air, Vapor and Liquid Water .. . . . . . . . . . . . . . . . . . .... 223 13.1 The Temperature Is Lower than Critical Temperature Tc . . . . . . . . . 223 13.2 State Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 223

xii

Contents

13.3

13.4 13.5

13.6

13.7 13.8

13.9

13.10 13.11

13.12

13.13 13.14

13.15

13.16

Quantities Which Describe the Evolution and the Thermal Heterogeneity . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.3.1 Classical and Generalized Deformation Velocities .. . . . 13.3.2 Thermal Heterogeneity . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Mass Balances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Equations of Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.5.1 The Power of the Interior Forces . . . .. . . . . . . . . . . . . . . . . . . . 13.5.2 The Power of the Acceleration Forces .. . . . . . . . . . . . . . . . . 13.5.3 Power of the Exterior Forces . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.5.4 The Equations of Motion . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Entropy Balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.6.1 The Energy Balance . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.6.2 The Second Law . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.6.3 The Free Energies .. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.6.4 The Entropy Balance . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . The Second Law: An Equivalent Formulation . . . . . . . . . . . . . . . . . . . . Pseudopotential of Dissipation .. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.8.1 Internal Constraints on the Velocities .. . . . . . . . . . . . . . . . . . 13.8.2 The Pseudopotential of Dissipation .. . . . . . . . . . . . . . . . . . . . Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.9.1 The Dissipative Internal Forces . . . . .. . . . . . . . . . . . . . . . . . . . 13.9.2 The Constitutive Laws . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.9.3 Consequences of the Constitutive Laws . . . . . . . . . . . . . . . . The Equations of the Predictive Theory . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.10.1 A Potential and the Free Enthalpy . .. . . . . . . . . . . . . . . . . . . . Air, Liquid Water and Vapor in an Homogeneous Cloud . . . . . . . . . 13.11.1 The Air, Liquid Water, Vapor Mixture .. . . . . . . . . . . . . . . . . 13.11.2 The Cloud Pressure . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.11.3 The Liquid Water . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.11.4 The Air, Vapor Mixture .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . A Schematic Example in Case There Is No Air . . . . . . . . . . . . . . . . . . . 13.12.1 The Specific Entropies and the Specific Phase Change Latent Heat. . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Case Where Vapor, Liquid Water and Air Velocities Are Equal .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Case Where Air Velocity Is Null, Vapor and Liquid Velocities Are Equal and Small . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.14.1 The Example . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.14.2 The Equations When the Temperature Is Known . . . . . . The Temperature Is Larger than Critical Temperature Tc . . . . . . . . . 13.15.1 An Homogeneous Cloud . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13.15.2 The Example: The Different TemperatureSpecific Volume Domains . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . Solid–Solid and Solid–Gas Phase Changes. . . .. . . . . . . . . . . . . . . . . . . .

224 225 225 226 227 227 228 229 230 231 231 233 234 238 240 240 241 241 242 242 243 244 245 249 251 251 252 253 255 256 259 260 261 262 264 265 266 266 268

Contents

xiii

14 Conclusion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 269 A

Some Elements of Convex Analysis . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . A.1 Convex Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . A.2 Convex Functions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . A.2.1 Examples of Convex Functions . . . . .. . . . . . . . . . . . . . . . . . . . A.3 Linear Spaces in Duality.. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . A.3.1 Examples of Linear Spaces in Duality .. . . . . . . . . . . . . . . . . A.4 Subgradients and Subdifferential Set of Convex Functions.. . . . . . A.4.1 Two Properties of the Subdifferential Set. . . . . . . . . . . . . . . A.4.2 Examples of Subdifferential Sets . . .. . . . . . . . . . . . . . . . . . . . A.5 Dual Functions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . A.5.1 The Internal Energy, Potential G and Free Enthalpy .. . A.6 Concave Functions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . A.6.1 Example of Concave Functions . . . . .. . . . . . . . . . . . . . . . . . . .

271 271 271 272 273 273 274 275 276 278 278 279 280

B

The Small Perturbation Assumption . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.1 Assumptions on the Displacement and on Mechanical Quantities.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.1.1 The Surface Normal Vector . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.2 Assumptions on the Exterior Forces: The Choice of the Small Perturbation Equations . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.3 The Mass Balance.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.4 The Equations of Motion . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.4.1 The Equation for the Macroscopic Motion . . . . . . . . . . . . . B.4.2 The Equation for the Microscopic Motion: The Classical Case . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.4.3 The Equation for the Microscopic Motion: The Non Classical Case . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.5 The Entropy Balance . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.6 The Constitutive Laws . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.7 Example 1: The Stefan Problem . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.8 Example 2: The Damage Predictive Theory .. .. . . . . . . . . . . . . . . . . . . . B.8.1 Some Numbers for the Equation for the Microscopic Motion . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.9 A Practical Rule .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . B.10 Advantages and Disadvantages of the Small Perturbation Assumption . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .

281 281 284 285 285 286 286 287 288 289 289 290 291 292 293 293

References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 295

•

Chapter 1

Introduction

A model is a theory which predicts the evolution of a structure. It does not explain the underlying physical phenomenon, it only predicts what is going to happen: physics explains, mechanics predicts the motion. A model is schematic and limited to some aspects of the phenomenon which occurs. In this point of view, the scientist or the engineer has an important and major role. He chooses the quantities which are to be predicted. These quantities are the state quantities: they characterize the equilibrium, i.e., the structure is at an equilibrium if the state quantities are constant with respect to the time. Let us stress that the notion of equilibrium is subjective. In the sequel we consider phase change occurring at the engineering level, i.e., at the macroscopic level: solid–liquid phase changes, damage phenomenon, phase change involving temperature discontinuities,... Many other phase change phenomena are relevant to this theory: solid–solid phase changes as in shape memory alloys, phase changes in porous medium, vapor-liquid phase changes,... The classical ice–water phase change involves microscopic motions which have macroscopic effects. In shape-memory alloys the solid–solid phase change involves also microscopic motions which are responsible for phase twinning. When damaging structures, microscopic motions break the links which are responsible for the cohesion of the material. We think that all these microscopic motions have to be taken into account in the macroscopic predictive theories. The basic idea we have developed is to account for the power of the microscopic motions in the power of the interior forces. Thus we modify the expression of the power of the interior forces and assume that it depends on the volume fraction rate, micro-voids volume fractions rate, damage rate,... which are clearly related to the microscopic motions. Furthermore we assume that it depends also on the gradient of these rates to account for local microscopic interactions. The consequences of this assumption give the basic equations of motions, one for the macroscopic motions and another one for the microscopic motions. They may be applied in numerous predictive theories: phase change problems, damage, shape memory alloys,... Moreover when phase change occurs, temperatures may be discontinuous, for instance when warm rain falls on frozen soil. M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 1, © Springer-Verlag Berlin Heidelberg 2012

1

2

1 Introduction

The notes are organized as follows: a scheme for predictive theories is given in a rather general framework using state quantities related to applications: after choosing the state quantities (Chap. 2), the equations which relate them are given: the balance laws (Sects. 3.1, 3.2 and 3.2.4) and the constitutive laws (Sect. 3.3). Their status is slightly different. The balance laws have a large scope even if they may be carefully modified to adapt to the situation under consideration. For instance, the classical equations of motion are not intangible and they may be completed to deal with enhanced description of motion as mentioned above, [112]. On the contrary the constitutive laws are very specific and in general apply only to the model under consideration. Then examples are described: numerous solid–liquid phase changes (Chap. 4), solid–solid phase changes in shape memory alloys (Chap. 5), damage theories (Chap. 6), contact with adhesion (Chap. 7), the coupling of volume damage and surface adhesion (Chap. 8) and phase change with temperature discontinuities (Chap. 9). In the damage and adhesion theories, we investigate what occurs when the distinction between macroscopic and microscopic motions, which is the very base of the two equations of motion, is not so obvious. We show that effects of vanishing macroscopic motions are taken into account by the equation describing the microscopic motion. Thus the two types of equations of motion may be connected and may be not totally independent. The phase changes produced by the thermal effects of collisions are investigated in Chaps. 10 and 11. It is predicted the behaviour of warm rain which falls on a frozen ground. The previous phase change temperatures are fixed, for instance the ice water phase change occurs at 0◦ C. But there are phase change temperatures which depend on a physical quantity: for instance, the pressure. In Chap. 12, we investigate liquid vapor phase change, for instance boiling water. This phase change depends on one parameter: when there is phase change occurring on a free surface which separates liquid from vapor, the state quantities satisfy a relationship, for instance, the temperature is a function either of the pressure or of the specific volume. But it is also possible that liquid and vapor coexist in a volume. This is the case in mist, fog and clouds. The air, vapor, liquid mixture is investigated in Chap. 13. The set of state quantities may depend on the temperature: when the temperature is lower than a critical temperature Tc , there are two phases vapor and liquid with two volume fractions which are state quantities and when the temperature is larger than the critical temperature, there is only one phase, the gas with only one volume fraction. In the first Appendix A some useful properties of convex analysis are given. In the second Appendix B the small perturbation assumption is investigated.

Chapter 2

The State Quantities and the Quantities Describing the Evolution

2.1 The State Quantities As already said, the state quantities are the quantities chosen by the scientist or the engineer to describe a phenomenon he wants to predict. The state quantities define the equilibrium: when the state quantities remain constant with respect to the time, the phenomenon is at an equilibrium. The notion of equilibrium is subjective and depends on the sophistication of the modelling, i.e., on the choice of the state variables. In order to be specific and apply the results to phase change problems, we choose the state quantities to be: • T , the absolute temperature which is clearly important. • β , the volume fraction of a phase of a two phase material. For instance, the liquid water volume fraction in ice–water phase change. This quantity is important to make more precise the description of the mixture of the different phases. • gradβ , the gradient of the volume fraction to account for local interactions and to investigate sophisticated phase change phenomena. • ε , the small deformation which is chosen because it may be useful to take into account the deformation of the materials. For the sake of simplicity the small perturbation assumption (see Appendix B) is made when needed. Thus the set of the states quantities we have chosen is E = (T, ε , β , gradβ ) . For the sake of simplicity, we assume the material density ρ is constant with respect to time in any of its phases (see Sect. 4.6 in case densities differ in the material phases). The case it is not constant when dealing with clouds is investigated in Chap. 13.

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 2, © Springer-Verlag Berlin Heidelberg 2012

3

4

2 The State Quantities and the Quantities Describing the Evolution

2.2 The Quantities Which Describe the Evolution and the Thermal Heterogeneity Besides the equilibrium, the engineer or the scientist has to describes evolutions. The velocities of the state quantities are important but it may be useful to have other quantities depending on the present and the past to describe the evolution. This is the case of the gradient of the temperature, gradT, which accounts for the thermal heterogeneities. The set of quantities describing the evolution we use is 

δE =

 dε dβ dβ , , grad , gradT . dt dt dt

Chapter 3

The Basic Laws of Mechanics

The laws we need to build models for phase change problems are the principle of virtual power to get the equations of motion and the laws of thermodynamics to deal with thermal actions. For the sake of simplicity, we assume the small perturbation assumption (see Appendix B) after deriving the equations of motion. Within this assumption, the mass balance which describes the density evolution has not a major role, except for phase change with voids (Sect. 4.5), phase change with different densities (Sect. 4.6).

3.1 The Principle of Virtual Power The equations of motion are derived with the principle of virtual power (a power is a duality pairing in mathematical terms). The power of the interior forces is chosen to depend on the regular strain rates D(U) (U is the macroscopic actual velocity and Di , j (U) = 1/2(Ui, j +U j ,i )), and also on dβ /dt and grad(dβ /dt), where β stands for some internal quantity, for instance a phase volume fraction or microvoids volume fraction. These latter quantities are clearly related to the microscopic motions which intervene in the evolution of the material micro-structure, for instance in the evolution of micro-voids. The gradient of the volume fraction is introduced to take into account local interactions or the influence of a material point on its neighbourhood. By stretch of language, derivative dβ /dt is referred as microscopic velocity. The virtual power of the interior forces we choose, [112], is Pint (V, γ ) = −

 Ω

σ : D(V)d Ω −

 Ω

(Bγ + H · gradγ ) d Ω ,

(3.1)

where V, γ are macroscopic and microscopic virtual velocities and Ω is the domain with boundary ∂ Ω , occupied by the material. Two new non classical interior

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 3, © Springer-Verlag Berlin Heidelberg 2012

5

6

3 The Basic Laws of Mechanics

forces appear, B, the interior microscopic work, and H, the microscopic work flux vector. The tensor σ is the stress tensor which is symmetric in this setting (see Remark 3.3 below). Let us note that the expression of the volume density of virtual power gives the quantities which are to be related by constitutive laws and to be measured in experiments. In some case, the generalized forces and the generalized strain rates to be related are not so easy to identify and in this perspective the choice of the power of the internal forces is an important element of the predictive theory, [112]. Remark 3.1. The scalar product of 3 × 3 matrices is

σ : D = σi j Di j , see Appendix A. Remark 3.2. The power Pint (V, γ ) has to be null for any rigid body motion, [60, 131,132,197]. A rigid body motion is a motion such that the distance of the material points remains constant, i.e., a rigid body translation or a rigid body rotation. At the microscopic level, in such a motion there is no phase change which requires deformation of the microscopic structure. Thus volume fraction β remains constant, d β /dt = 0. It results that rigid body motions are defined by D(V) = 0, γ = 0 and that Pint (V, γ ) = 0 for such motions. The virtual power of the exterior forces is Pext (V, γ ) =

 Ω

f · Vd Ω +

 ∂Ω



g · VdΓ +

Ω

Aγ d Ω +

 ∂Ω

aγ d Γ ,

where f is the volume exterior force, g the surface exterior force, A and a are respectively the volume and surface exterior sources of microscopic work. Sources of microscopic work can be produced by chemical, electrical, or radiative actions which break the links inside a material, steel or concrete for instance, without macroscopic deformations, [21, 24, 112, 113, 120, 121, 175]. The virtual power of the acceleration forces is Pacc (V, γ ) =

 Ω

ρ

dU · Vd Ω + dt

 Ω

ρ0

d2 β γ dΩ , dt 2

where ρ is the density of the solid and ρ0 is proportional to the density of the microscopic links. The principle of virtual power ∀ V, γ , Pacc (V, γ ) = Pint (V, γ ) + Pext (V, γ ),

3.1 The Principle of Virtual Power

7

gives easily the equations of motion

ρ ρ0

dU − divσ = f, in Ω , σ N = g, on ∂ Ω , dt

d2 β − divH + B = A, in Ω , H · N = a, on ∂ Ω . dt 2

(3.2)

The second equation of motion (3.2) is new. It accounts for the microscopic motions. Both the partial differential equation and the boundary condition have a precise physical meaning: they describe how work is provided to the structure without macroscopic motion. In the sequel, we consider mostly quasi-static evolutions. Results involving the accelerations are reported in [49, 123]. Remark 3.3. The power of the internal forces is null for any rigid body translation velocity because of the Galilean relativity. It is null for any rigid body angular velocity because of the angular momentum balance assuming no external torque and resulting in this setting in the symmetry of the stress tensor. Note that those two properties are quite different. This presentation is very attractive due to its simplicity and its large scope. Its only weakness is to forbids external torques but in small deformation engineering it is rare that applications involve torques. A more general setting is Pint (V, γ ) = − Pext (V, γ ) =



 Ω

Ω

σ : gradVdΩ −

1 m : ω (V)d Ω + 2

 

Ω

Ω

(Bγ + H · gradγ ) dΩ, f · Vd Ω +

 ∂Ω

g · VdΓ +

 Ω

Aγ d Ω +

 ∂Ω

aγ dΓ ,

where 1 ω (V) = (gradV − (gradV)T ), 2 the virtual angular velocity is the antisymmetric part of the virtual velocity gradient and antisymmetric matrix m is an external torque (matrix (gradV)T is the transposed matrix of matrix (gradV)). The principle of virtual power gives equations of motion (3.2) where stress σ is not symmetric and relationship

σ − σ T = m, which is the angular momentum balance. It results antisymmetric part of the stress is given by the equation of motion and its symmetric part is going to be given by a constitutive law. In case there is no external torque, m = 0, we recover the presentation we have chosen with power of the internal forces given by (3.1) with symmetric stress σ .

8

3 The Basic Laws of Mechanics

3.2 The Laws of Thermodynamics Besides the description of motion, the models involve thermal and energetic phenomena. Thus the laws of thermodynamics intervene: the first law, often called the energy balance, together with the second law, see [112] for a detailed presentation.

3.2.1 The Energy Balance The energy balance coupled to the principle of virtual power gives the following balance law, [112] de dβ dβ + divTQ = σ : D(U) + B + H · grad + RT, in Ω, dt dt dt −T Q · N = T π , on ∂ Ω ,

(3.3)

where T is the absolute temperature, e is the volumic internal energy, Q is the entropy flux vector (T Q = q is the heat flux vector), R the volumic exterior rate of entropy production (T R = r is the volumic exterior rate of heat production) and π the surfacic rate of entropy provided to the material (T π = ϖ is the surfacic rate of heat provided to the material). By denoting F meca = (σ , B, H, 0) , energy balance (3.3) reads de + divTQ = F meca · δ E + RT, in Ω , dt −T Q · N = T π , on ∂ Ω .

(3.4)

Remark 3.4. The opposite of the power of the interior forces F meca · δ E = σ : D(U) + B

dβ dβ + H · grad , dt dt

is often neglected within the small perturbation assumption (see Appendix B). The energy balance becomes de + divq = r, in Ω , dt −q · N = ϖ , on ∂ Ω . It is the classical energy balance.

3.2 The Laws of Thermodynamics

9

3.2.2 The Second Law of Thermodynamics It is T > 0,

(3.5)

ds + divQ ≥ R, in Ω , dt

(3.6)

where s is the entropy.

3.2.3 The Free Energy Let us recall a physical property: the internal energy is a convex function of entropy, [132]. By using convex analysis (see Appendix A), we may define the dual function of the internal energy, the free energy −Ψ (T, ε , β , gradβ ) = sup {xT − e(x, ε , β , gradβ )} , x∈R

giving e(s, ε , β , gradβ ) = Ψ(T, ε , β , gradβ ) + Ts,

(3.7)

and s=−

∂Ψ ∂e (T, ε , β , gradβ ), T = (s, ε , β , gradβ ). ∂T ∂s

(3.8)

Function T → Ψ (T, ε , β , gradβ ) is concave. This concavity property is a physical property because it results from convexity of function s → e(s, ε , β , gradβ ), [132]. In the sequel, we choose the free energy as the basic physical potential. In case we need internal energy, we define it by the previous relationship, (3.7).

3.2.4 Equivalent Laws of Thermodynamics The laws of thermodynamics (3.3), (3.5), (3.6) involve two partial differential equations. They have an equivalent formulation involving only one partial differential equation, the entropy balance law, and two inequalities. In order to derive this

10

3 The Basic Laws of Mechanics

balance law, the internal forces are split between dissipative interior forces indexed by d and non dissipative interior forces indexed by nd

σ = σ d + σ nd , B = Bd + Bnd , H = Hd + Hnd .

(3.9)

The non dissipative interior forces are defined with the free energy, Ψ (E) = Ψ (T, ε , β , gradβ ), which gathers all the static properties of the material

∂Ψ , ∂T ∂Ψ ∂Ψ . = , Hnd = ∂β ∂ (gradβ )

s=−

∂Ψ nd ,B ∂ε

σ nd =

(3.10)

By replacing internal energy in energy balance equation (3.3) by its value given by (3.7) and using relationship (3.5), we get equations equivalent to (3.3), (3.5) and (3.6), [41] ds 1 + divQ = R + dt T



 dβ σ : D(U) + B + H · grad − Q · gradT , in Ω , dt dt d

d dβ

d

−Q · N = π , on ∂ Ω , T > 0,   dβ 1 d d dβ d + H · grad − Q · gradT ≥ 0. σ : D(U) + B D= T dt dt

(3.11)

Let us note that only one partial differential equation is involved in this equivalent formulation.

3.2.5 The Entropy Balance Equivalent to the Energy Balance The first two relationships of (3.11) are the entropy balance ds + divQ = R + D, in Ω , dt −Q · N = π , on ∂ Ω ,

(3.12)

which is equivalent to the energy balance. The entropy balance shows that the variation of the entropy results from an exterior source, R and from an interior one, D. In terms of physics, this interior entropy source may be experimented when rubbing one’s hands. Rubbing produces heat but it never produces cold. This property is T D = Rd · U ≥ 0, where Rd is the contact reaction and U is the hands sliding velocity.

3.3 The Constitutive Laws

11

3.2.6 An Equivalent Second Law of Thermodynamics The two other relationships of (3.11) are T > 0,   dβ 1 d d dβ d + H · grad − Q · gradT ≥ 0. σ : D(U) + B D= T dt dt

(3.13)

The last relationship means that the interior production of entropy is non negative or that the dissipation is non negative. In some cases, these equations are more productive in terms of mechanics and mathematics than the classical energy balance equations, [41]. For instance, it may be straightforward to prove that the temperature is positive. We assume small perturbation (see Appendix B). It results that we may neglect the dissipation D=

1 T



σ d : D(U) + Bd

 dβ dβ + Hd · grad − Q · gradT , dt dt

in (3.12) which becomes ds + divQ = R, in Ω , dt −Q · N = π , on ∂ Ω .

(3.14)

New results on the case where the dissipation is not neglected may be found in [38, 39].

3.3 The Constitutive Laws They are relationships (3.9), (3.10) and relationships defining the dissipative interior forces which depend on the velocities dε /dt, dβ /dt, grad(dβ /dt), on the thermal heterogeneity gradT and possibly on other quantities χ which depend on the history of the material. The dissipative interior forces are defined by derivatives

σd =

∂Φ , ∂ (dε /dt)

Bd =

∂Φ , ∂ (dβ /dt)

12

3 The Basic Laws of Mechanics

Hd =

∂Φ , ∂ (grad(dβ /dt))

Q=−

∂Φ , ∂ (gradT)

(3.15)

where 

Φ

d ε dβ dβ , , grad , gradT, χ dt dt dt

 = Φ(δ E, χ ),

is a pseudopotential of dissipation introduced by Jean Jacques Moreau [130, 139, 171]: it is a function of δ E which is convex with respect to δ E, non negative, with value 0 for δ E = 0. In a more general setting, relationships (3.15) are 

   d ε dβ dβ σ , B , H , −Q ∈ ∂ Φ , , grad , gradT, χ = ∂ Φ (δ E, χ ), dt dt dt d

d

d

(3.16)

where ∂ Φ is the subdifferential set of pseudopotential of dissipation Φ with respect to δ E, [112]. These constitutive laws ensure that relationship (3.13) is satisfied: Theorem 3.1. If T > 0 and Φ (δ E, χ ) is a pseudopotential of dissipation, i.e., function δ E → Φ (δ E, χ ) is convex, non negative, with value 0 at the origin, Φ (0, χ ) = 0, then constitutive laws (3.16) are such that second law relationship (3.13) is satisfied. Proof. Due to definition of subdifferential (see Appendix A), we have  0 − σd :

dε dt

 + (0 − Bd )

dβ dβ + (0 − Hd ) · grad + (0 − (−Q)) · gradT dt dt   dβ dε dβ , , grad , gradT, E ≤ Φ(0, E). +Φ dt dt dt

Because Φ (0, E) = 0, we have 

Φ

 dβ dε dβ dβ dε d β , , grad , gradT, E ≤ σ d : + Bd + Hd · grad − Q · gradT, dt dt dt dt dt dt

which gives relationship (3.13) because Φ (δ E, χ ) ≥ 0 and T > 0.

 

This way to define the dissipative forces is not the more general but its scope is very large and sufficient in many situations. Experiments are of paramount importance for the choices of Ψ and Φ . The derivatives of these functions are the quantities which can be measured with experiments. There are some empirical rules related to physical properties to choose the two functions.

3.4 Internal Constraints

13

The choice of the state quantities followed by the choice of Ψ and Φ are the crucial and delicate steps in modelling. The imagination and the creative ability, even the artistic ability, of the engineer or scientist may appear when establishing the closed forms of functions Φ and Ψ , resulting in elegant and useful predictive theories. Remark 3.5. The main practical problem which is not described by a pseudopotential of dissipation is the Coulomb friction law. New results on constitutive laws where gradT intervene are given in [177]. Remark 3.6. The mass balance is not taken into account. We have assumed for the sake of simplicity that the density is constant. If the mass balance is taken into account it result a pressure in the stress constitutive law as may be been seen in a smooth evolution described in Sect. 4.5 and in Chap. 13).

3.3.1 An Extra State Quantity A state quantity η of E different from T may be such that its velocity d η /dt does not intervene in the power of the internal forces. This case is investigated in [112]. The new equation related to η is Z d + Z nd = 0, with Z nd =

∂Ψ nd ∂Φ , ,Z = ∂η ∂ (dη /dt)

in case Ψ and Φ are smooth functions. In case they are not, the partial derivatives are replaced by the components of the subgradients of Ψ and Φ . In these notes, there are examples in the theory of shape memory alloys, see Sect. 5.5.2, where η = δ , the degree of education, and in the predictive theory of clouds in Chap. 13.

3.4 Internal Constraints The quantities of δ E which describe the evolution, are elements of linear spaces δ V , because they are velocities and gradT is an element of a linear space, δ E ∈ δ V . Their actual values are often not the entire linear space but a subset δ C ⊂δ V . It is said that the evolution quantities satisfy internal constraints. For example, the material may be incompressible   dε = 0, divU = tr dt

14

3 The Basic Laws of Mechanics

or a mass balance has to be satisfied

∂ (ρ1 β1 + ρ2 β2 ) + ρ divU = 0, ∂t

(3.17)

or an evolution is irreversible dβ ≤ 0, dt

(3.18)

or an non interpenetration condition has to be satisfied U · N ≥ 0, where U is the velocity of a point sliding on a solid with exterior normal vector N. These properties are physical properties and they are not mathematical properties. Thus they have to be taken into account by the two functions which gather the mechanical properties, either Ψ or Φ . It is obvious that

δE ∈ δC ,

(3.19)

has to be taken into account with Φ . We use a very productive and general way to satisfy (3.19) having the pseudopotential of dissipation to involve the indicator function of δ C , Iδ C

Φ (δ E, χ ) = Φ (δ E, χ ) + Iδ C (δ E). A difficulty appears immediately due to the non differentiability of function Φ . It is possible to overcome it by defining generalized derivatives. In many cases the set δ C is convex (this is the case for the examples) allowing to use subgradients as generalized derivatives. A short presentation of convex analysis is given in Appendix A. There are also internal constraints on the state quantities E. We apply the same idea. If the actual internal quantities have to satisfy E ∈C ⊂V, where C is a subset of space V , we choose the free energy satisfying

Ψ (E) = Ψ (E) + IC (E). In many cases, space V is linear but this is not always the case, [112]. In many practical situations, C is a convex subset of linear space V (for instance, if the small perturbation assumption is assumed), allowing to have subgradients as generalized derivatives. Examples of internal constraints on the state quantities are

β ∈ [0, 1],

(3.20)

3.4 Internal Constraints

15

implying that β is a volume fraction, (β1 , β2 ) ∈ K, the volume fractions (β1 , β2 ) of a mixture satisfy some relationship. If K is K = {(β1 , β2 ) |0 ≤ β1 ≤ 1, 0 ≤ β2 ≤ 1, β1 + β2 ∈ [0, 1] }, voids may appear in the mixture (see Sect. 4.5). The temperature T satisfy the internal constraint T ≥ 0.

(3.21)

For the sake of simplicity, we assume that (3.21) is always satisfied and we do not take it into account. The case T = 0 is investigated in [112]. We prove that solutions of most of the problems we solve below, satisfy this internal constraint if the initial temperature is positive.

3.4.1 Constitutive Laws in Case There Are Internal Constraints on the State Quantities For the sake of simplicity, we assume internal constraint (3.20) is satisfied by β and that trε = 0. The free energy becomes

Ψ (T, ε , β , gradβ ) + I(β ) + I0(trε ), where I and I0 are the indicator functions of segment [0, 1] and of the origin 0 of R, see Appendix A. We define the non dissipative reaction forces Breac ∈ ∂ I(β ), −preac ∈ ∂ I0 (trε ), σ reac = −preac 1, where ∂ I(β ) and ∂ I0 (trε ) are the subdifferential sets of the indicator functions I and I0 , see Appendix A. These constitutive laws imply that the subdifferential sets ∂ I(β ) and ∂ I0 (trε ) are not empty. Thus due to Theorem A.2 of Appendix A that I(β ) = 0 and I0 (trε ) = 0. Thus β ∈ [0, 1] and trε = 0. We replace definitions (3.9) by

σ = σ d + σ reac + σ nd , B = Bd + Breac + Bnd , H = Hd + Hnd .

(3.22)

16

3 The Basic Laws of Mechanics

It results entropy balance (3.12) where we keep definition (3.10), becomes   dβ dβ ds 1 σ d : D(U) + Bd + divQ = R + + Hd · grad − Q · gradT dt T dt dt   dβ 1 , in Ω , σ reac : D(U) + Breac + T dt −Q · N = π , on ∂ Ω .

(3.23)

We let 

 dβ + H · grad − Q · gradT , σ : D(U) + B dt dt   1 reac reac reac dβ = σ : D(U) + B D . T dt 1 D= T

d

d dβ

d

Quantity D is the dissipation and D reac is the reactions dissipation. The entropy balance is ds + divQ = R + D + D reac , in Ω , dt −Q · N = π , on ∂ Ω , where R is the exterior entropy source, D is the interior dissipative source and D reac is the non dissipative reaction source. The definition of the dissipative forces are still given by (3.16) if there are internal constraints on the velocities of δ E, for instance (3.17) or (3.18).     d ε dβ dβ d d d σ , B , H , −Q ∈ ∂ Φ , , grad , gradT, χ = ∂ Φ (δ E, χ ). dt dt dt Remark 3.7. The reactions σ reac and Breac are subgradients of the free energy thus they may be said non dissipative as the smooth derivatives σ nd , Bnd and Hnd which do not intervene in the entropy production. But these reactions may have a non null power which may intervene in the entropy production. We choose to keep the non dissipative qualifying adjective because the power of these reactions is null in any smooth evolution as shown in the following section.

3.4.1.1 The Power of the Non Dissipative Reactions Forces Let us note that

σ reac : D(U) = −preactr(D(U)) = −preac

d (trε ) = 0. dt

3.4 Internal Constraints

17

This property is general: in case the constraint is to belong to a linear subspace, the power of the non dissipative reaction force is null. In relationship (3.23), the velocities, for instance dβ /dt may be discontinuous with respect to time. Based on the causality principle [112], we choose to have the left derivative which depends on the past, in the formula dl β β (t) − β (t − Δ t) = lim . dt Δ t→0,Δ t>0 Δt

(3.24)

Let us investigate the power of reaction Breac : • In case 0 < β (t, x) < 1, we have 0 = Breac (t, x) ∈ ∂ I(β (t, x)) and Breac (t, x)

dl β (t, x) = 0. dt

• In case β (τ , x) = 1 in some interval [t0 ,t] with t0 < t, we have (d l β /dt)(t, x) = 0 and Breac (t, x)

dl β (t, x) = 0. dt

We have the same result in case β (τ , x) = 0 in some interval [t0 ,t] with t0 < t. • In case β (t, x) = 1 and β (τ , x) < 1 in some interval [t0 ,t] and (d l β /dt)(t, x) > 0, we have Breac (t, x) ∈ ∂ I(1) = R+ . If Breac (t, x) > 0, we have Breac (t, x)

dl β (t, x) > 0. dt

Breac (t, x)

dl β (t, x) = 0. dt

If Breac (t, x) = 0, we have

Let us note that Breac (t, x) = 0, if function τ → Breac (τ , x) is continuous at time t because Breac (τ , x) = 0 for t0 ≤ τ < t. Analogous result hold in case β (t, x) = 0 and β (τ , x) > 0 in some interval [t0 ,t]. We conclude that the power of reaction force Breac is null if it is a continuous function of time. In this situation we say that the constraints are workless (les liaisons sont parfaites in French). Let us note that the reaction power is always non negative Theorem 3.2. If temperature is positive and if the velocities are the left derivatives, we have D reac ≥ 0.

18

3 The Basic Laws of Mechanics

If the left derivatives are equal to the right derivatives, i.e., the state quantities the derivatives of which are involved in D reac are differentiable, we have D reac = 0. Proof. Proof is based on convex analysis, see [112].

 

Remark 3.8. A precise derivation of this point of view is given in [112] together with examples of situations where δ V is not a linear space and with an investigation of the internal constraint (3.21) on the temperature. In [112], it is also proved that when the velocities are discontinuous with respect to time, the left derivatives, are the quantities which appear in the basic laws of mechanics in agreement with the causality principle. Let us note that if the right derivative is chosen, it is not possible to prove that the second law is satisfied. In the sequel, the time derivatives are left derivatives. For the sake of simplicity, we write dβ /dt instead of dl β /dt. Right derivatives appear in collision theories but they sum up the very sophisticated phenomena which occur in a collision. With this point of view, they depend on the past in agreement with the causality principle, [114]. Remark 3.9. The power of the reaction is always non negative. We say that the evolution is smooth in case it is workless, i.e., it is null. This is the case, if either the reaction is a continuous function of time or the generalized velocity of deformation is continuous with respect to time. In our example, this condition is equivalent to the differentiability of the state quantity β . In terms of mathematics, it is often sufficient that the properties are satisfied almost everywhere with respect to x and t. In the sequel, we consider evolutions which are smooth, i.e., evolutions such that D reac = 0.

(3.25)

The case where the evolutions are not smooth is investigated when dealing with discontinuities, in Sects. 3.5 and 3.6 of this chapter and in Chaps. 9, 10 and 11 dedicated to discontinuities of temperature T and volume fraction β and to discontinuities of velocity U occurring in collisions. Remark 3.10. As already said, relationship (3.25) is chosen to define a smooth evolution. This is the case in rigid bodies mechanics where a workless constraint is defined by the fact its power is null. When solving a problem, one may verify a posteriori that the non dissipative reactions are workless, i.e., D reac = 0. It may also occur that it is impossible to solve the problem, i.e., to find a smooth solution. In this situation, a non smooth solution satisfying D reac > 0, may be found. The reaction dissipation is positive, i.e., the reactions are not workless (non parfaites in French). This is the case in non smooth evolutions, for instance in collisions as described in Chaps. 10, 11 and in [114].

3.5 The Basic Laws of Mechanics on a Discontinuity Surface

19

3.4.2 Some Notations In case there are many state quantities and velocities as in Chap. 13, it is useful to introduce some vectors gathering the dissipative internal forces F d , the non dissipative reactions to the internal constraints on the state quantities F reac and the non dissipative internal forces F nd  E = (T, ε , β , gradβ , η ) , δ E =

 dβ dη d ε dβ , , grad , , gradT , dt dt dt dt

F = (σ , B, H, 0, −Q), F = F d + F reac + F nd ,   F d = σ d , Bd , H, Z d , −Q , F reac = (−preac 1, Breac , 0, 0, 0) ,   F nd = σ nd , Bnd , H, Z nd , 0 .

(3.26)

We have  1 d F ·δE , T 1 D reac = (F reac · δ E) . T D=

Let us recall that a smooth evolution is characterized by D reac = 0.

3.5 The Basic Laws of Mechanics on a Discontinuity Surface The quantities which appear in the preceding laws have to satisfy some mathematical smoothness properties in order to define the derivatives. There are cases where these quantities are discontinuous and the results are no longer valid. Of course, the basic laws apply in these circumstances but they have to be adapted to take the discontinuities into account, [111, 112]. The discontinuities can be on a surface, which is often called a free boundary, when solving a problem in R3 (on a line when in R2 , on a point when in R). We assume that the specific volume 1/ρ is one of the state quantities E. We no longer assume small perturbations. Of course, results also make sense within the small perturbation assumption. When velocities U, β˙ = dβ /dt, temperature T , entropy s = ρ s, internal energy e = ρ e, entropy flux vector Q or interior forces σ and H are discontinuous with respect to x,t on a surface Sx,t of R3 × R, the computations which give the basic relationships are no longer valid. In this section, notation [A] = A2 − A1 denotes the discontinuity of the quantity A, m is the mass flux (m = ρ v, where v = Vrel · N is the material normal velocity with

20

3 The Basic Laws of Mechanics

respect to the surface, (see formula (3.31) below) and Q is the entropy received through contact actions by the two domains which are on the two sides of the discontinuity surface Γ = Sx,t ∩ (R3 × {t}). The indices 1 and 2 refer to the two sides of the discontinuity surface, the normal vector N is directed from side 1 toward side 2. Remark 3.11. In this section, Γ has dimension 2. When Γ = Sx,t ∩ (R3 × {t}) has dimension 3, i.e., when the domain where there is a discontinuity at time t has dimension 3 and the discontinuity is with respect to time, another analysis is needed. This situation is investigated in the following section. The second law of thermodynamics on the discontinuity surface Γ with unit normal vector N in R3 , is m [ s ] ≥ [Q],

(3.27)

→ − where sis the specific entropy and Q = −Q. N is the entropy flow through surface Γ , T Q is the heat flow at temperature T . The inequality depends only on the difference [Q] = Q2 − Q1 , the entropy received by the surface. The distribution of [Q] between the received entropies from the two sides, −Q1 from side 1 and Q2 from side 2, results from the constitutive laws which are established in the sequel. We denote T = σ N the force applied on the discontinuity surface Sx with normal vector N. Different average values are defined by T=

T1 + T2 Q1 + Q2 U1 + U2 T1 + T2 A1 + A2 ,Q= ,U= ,T= ,A= , 2 2 2 2 2

where A is the average of a quantity A on the discontinuity surface. Let us recall that the temperatures T are positive because they are the traces on the discontinuity surface of the volume temperatures which are positive. We assume that the average temperature T is positive. The particular case where T = 0 is investigated in [112]. The equation of motion are m[U] = [T],

(3.28)

and [H · N] = 0. The energy balance is 

1 2 m e+ U − [T · U] − [H · Nβ˙ ] = [T Q] = T [Q] + Q[T ], 2

(3.29)

where we recall that e is the specific internal energy. Relationship (3.28) gives [T · U] = T · [U] + U · [T] = T.[U] + mU.[U] = T.[U] +

m 2 [U ]. 2

(3.30)

3.5 The Basic Laws of Mechanics on a Discontinuity Surface

21

Note that due to the mass balance [ρ (U − W)] · N = 0,  [UN ] = [U] · N = [Vrel ] · N = [VN ] = [v] =

 ρv 1 =m , ρ ρ

(3.31)

where, UN = U · N, Vrel = U − W is the velocity with respect to the discontinuity surface, whose velocity is W, v = VN = Vrel · N, the relative normal velocity. Let us also define V = (Vrel2 + Vrel1 )/2, the average relative velocity. With the relationships (3.30) and (3.31), the energy balance (3.29) gives    1 m [ e] − TN − TT · [UT ] − H · N[β˙ ] − T [Q] − Q[T ] = 0, ρ

(3.32)

where T N = T · N, TT = T − T N N and UT = U − UN N. In relationship (3.32), the  is the specific free energy to get  + T s, where Ψ internal energy e is replaced by Ψ     + T s] − T N 1 m [Ψ − TT · [UT ] − H · N[β˙ ] − T [Q] − Q[T ] = 0, ρ or  

1  T {[m[ m + TT · [UT ] + H · N[β˙ ] + Q[T ]} = 0. s ] − [Q]} − −([Ψ ] + s[T ] − T N ρ (3.33)  is replaced  ] + s[T ] − T N [1/ρ ]) is perfectly defined. Actually, if Ψ Quantity −([Ψ  +CT + A, (A and C are constants, let us recall that the free energy is defined up by Ψ  ] + s[T ] is replaced by ([Ψ  ] + C [T ]) + to an affine function of the temperature), [Ψ  ( s − C)[T ] = [Ψ ] + s[T ]. In order to use equality (3.33), the state quantities and the actual or virtual evolution of the material are to be defined. We decide to characterize the state of the material by the quantities (ε , 1/ρ , β , gradβ , T) on the two sides of the discontinuity surface   1 1 E = ε1 , , β1 , gradβ1 , T1 , ε2 , , β2 , gradβ2 , T2 , ρ1 ρ2 and its evolution by the velocities and discontinuities   Δ E˜ = VN2 = Vrel2 · N,VN1 = Vrel1 · N, [T ], [VT ], [β˙ ] , which describe the way the state of the material evolves when crossing the discontinuity surface and the way the discontinuity line moves. Note that this set is equivalent to ((VN2 + VN1 )/2, [T ], [V], [β˙ ]) or to ((VN2 + VN1 )/2, [T ], [U], [β˙ ]) since [V] = [U]. This last choice is coherent with the choice of ε (U), gradβ˙ and gradT as quantities describing the evolution when there is no discontinuity. Actually [U],

22

3 The Basic Laws of Mechanics

[β˙ ] and [T ] are the non-smooth part of the space derivatives which appear in δ E, (ε (U), gradβ˙ , gradT). The average normal velocity is added in order to describe the evolution of the discontinuity line. One can say that the constitutive laws depend on (ε (U), gradβ˙ , gradT) in the distributive sense. Remark 3.12. More details are given in [112]. Following [112], the choice of Δ E˜ is in agreement with the characterizations of the evolution by left derivatives. The quantities Δ E˜ are, of course, known, because the left derivatives are known in the smooth situation, as soon as the history of the material is known. The equations on a discontinuity surface give the initial velocities which depend on the past for the future evolution. The inverse of the densities ρ , the specific volumes, are state quantities. Thus we can choose ρ2VN2 and ρ1VN1 as elements equivalent to VN2 and VN1 , since the ˜ gives the same information as the set {E, (ρ2VN2 , ρ1VN1 , [T ], [VT ])}. set (E, Δ E) ˜ is Assuming the mass balance (3.31) is satisfied, we find that knowing (E, Δ E) equivalent to knowing (E, Δ E) with

Δ E = (m, [T ], [UT ], [β˙ ]). Thus we decide to characterize the evolution of the material by Δ E. Remark 3.13. The quantity [ρ ] is not chosen as an element of Δ E˜ or Δ E because the mass balance gives [ρ VN ] = [ρ ]V N + ρ [VN ] = 0. ˜ . Thus [ρ ] is a function of ρ [VN ] which is an element of a set equivalent to Δ E) Remark 3.14. In the smooth situation, vector gradT(x, t) is known as soon as the function T (x,t) is known. Nevertheless we have chosen T ∈ E and gradT ∈ δ E. On the discontinuity surface, to know T1 (x,t) and T2 (x,t) is equivalent to know function T (x,t) in the smooth situation, and to know [T (x,t)] is equivalent to know gradT(x, t) in the smooth situation. Thus the choices of T1 (x,t) ∈ E, T2 (x,t) ∈ E and [T (x,t)] ∈ Δ E are consistent. Vector F

dd

 

  1  , Q, TT , H · N , = − [Ψ ] + s[T ] − T N ρ

depends on the volume constitutive laws, i.e., on E. Relationship (3.33) becomes T {[m[s] − [Q]} − F dd · Δ E = 0, The constitutive laws on the discontinuity surface are relationships between F dd , Δ E and E such that F dd · Δ E ≥ 0,

3.5 The Basic Laws of Mechanics on a Discontinuity Surface

23

which is equivalent to the second law (3.27). Constitutive law for Q gives the distribution of the received entropy [Q] between the two sides of the discontinuity surface or between the temperatures T1 and T2 . In the same way, constitutive law for the force TT gives the distribution of the force [T] between the two sides of discontinuity surface. The constitutive laws are defined by a surface pseudopotential of dissipation Φ (Δ E, E) = Φ (m, [T ] , [UT ], [β˙ ], E) where E is a parameter. In case this potential is a smooth function, constitutive laws are  ] + s[T ] − T N [ 1 ]) = ∂ Φ , −([Ψ ρ ∂m Q=

∂Φ , ∂ [T ]

(3.34)

TT =

∂Φ , ∂ [UT ]

(3.35)

∂Φ . ∂ [β˙ ]

(3.36)

H·N =

In Sect. 4.1 and in Chaps. 9 and 12 down below, examples are given. In the classical situation, there are three constitutive laws on a discontinuity surface. Law (3.36) with

Φ (m, [T ] , [UT ]) = I0 ([T ]) + ϕ ([UT ]), gives [T ] = 0 and −TN = −

] [Ψ , [1/ρ ]

 /∂ (1/ρ ), which gives the pressure which is the discontinuous analogue to p = −∂ Ψ in a fluid where there is no dissipation with respect to ∂ ρ /∂ t or to divU due to the mass balance, (see for instance formula (13.46) below). Law (3.34) is analogous to law (3.15) which gives Fourier law. Law (3.35) is analogous to law (3.15) which gives the dissipative stresses depending on the strain rates. The tangential force TT is only dissipative. There is indeed no contribution of the free energy to its value. Note that the constitutive laws are actually causal: they characterize completely a material if its past is known, i.e., if its state E, its evolution Δ E depending on its history are known.  ], [T ], [1/ρ ] and m, Q and [T ], TT and [UT ], H ·N and [β˙ ] change sign Note that [Ψ when the orientation of the normal vector N is changed whereas s, T N , T , and [Q] do not change sign. We assume that the pseudo-potential Φ depends on the orientation and satisfies

Φ (m, [T ] , [UT ], [β˙ ], −N) = Φ (−m, − [T ] , −[UT ], −[β˙ ], N), vector N being understood as a state quantity.

24

3 The Basic Laws of Mechanics

3.6 The Basic Laws of Mechanics in a Discontinuity Volume Discontinuities can be at a time t in the whole volume. This is the case for collisions which are investigated in Chap. 11. We recall very rapidly the collision theory detailed in [112, 114]. The collisions being dissipative phenomena, they produce heat which intervene in the thermal evolution and may produce phase changes. In this section, we denote for function t → A(t) A=

A+ + A − , [A] = A+ − A− . 2

where A+ =

lim

Δ t→0,Δ t>0

A(t + Δ t), A− =

lim

Δ t→0,Δ t>0

A(t − Δ t),

are the values of quantity A after and before the discontinuity. Note that in an other context, values A+ and A− have been denoted right and left values, Ar and Al , see Sect. 3.4.1.1 and formula (3.24).

3.6.1 The Principle of Virtual Power and the Equations of Motion The virtual work of the interior forces, the interior percussion forces, is 



V+ + V− Tint (V, γ ) = − Σ : D 2 Ω

 dΩ −

 Ω

B p [γ ] + H p · grad [γ ] dΩ,

where V, γ are virtual velocities: V− , γ − are the velocity before the collision and V+ , γ + are the velocity after. It is denoted [X] = X + − X + , the discontinuity of quantity X. It appears percussion stresses Σ and percussion work B p and work flux vector H p . The virtual work of the acceleration forces is Tacc (V, γ ) =

 Ω

ρ [U] ·

V+ + V− dΩ . 2

Remark 3.15. The mass balance is [ρ ] = 0. Thus the densities are constant in the equations of motion. The incompressibility in collision of fluids is investigated in [85, 112, 115].

3.6 The Basic Laws of Mechanics in a Discontinuity Volume

25

The virtual work of the exterior forces is Text (V, γ ) =

 Ω

+

Fp ·

 Ω

V+ + V− dΩ + 2

A p [γ ] d Ω1 +

 ∂Ω

 ∂Ω

Gp ·

V+ + V− dΓ 2

a p [γ ] d Γ .

We assume that the surface exterior percussions G p are applied to the whole boundary of the solid. The F p are the volume exterior percussions. The A p and a p are the volume and surface percussion work provided by the exterior by electrical, radiative,... actions. The equations of motion results from the principle of virtual work ∀V, ∀γ , Tacc (V, γ ) = Tint (V, γ ) + Text (V, γ ). Different choices of the virtual velocities V and γ give

ρ [U] = divΣ + Fp , in Ω, ΣN = Gp , on ∂ Ω, −B p + divHp + Ap = 0, in Ω , H p N = a p , on ∂ Ω .

3.6.2 The Laws of Thermodynamics The first law is [E ] + [K ] = Text (U, β ) + C , where K is the kinetic energy, and C is the thermal impulse received by the structure. With the principle of virtual work where the velocities are the actual velocities, the first law gives [E ] = −Tint + C . The temperature may be discontinuous: we denote T − the temperature before the collision and T + the temperature after the collision. We assume that heat is received either at temperature T − or at temperature T + C=

 ∂Ω

 + +

− − − T Q p + T Q p · NdΓ + T + B+ + T − B − d Γ , Ω

where Q p is the impulsive entropy flux vector and B the impulsive entropy source. With E =

 Ω

ed Ω ,

26

3 The Basic Laws of Mechanics

we get  [e] = Σ : D

U+ + U− 2

 + B p [β ] + H p · grad[β ]

− − + + − − − div(T+ Q+ p + T Qp ) + T B + T B .

By using the Helmholtz relationship, e = Ψ + T s, we have 

 U+ + U− + B p [β ] + H p · grad[β ] [Ψ ] + s [T ] + T [s] = Σ : D 2 − div(TΣ Qp + [T] Δ Qp ) + TΣ (B) + [T] Δ (B) , (3.37) where a sum T + B + + T − B − = Σ (T B), is split in an other sum

Σ (T B) = T Σ (B) + [T ] Δ (B) , with

Σ (B) = B+ + B − , Δ (B) =

B+ − B− , 2

and T=

T+ + T− , [T ] = T + − T − . 2

The second law is [S ] =

 Ω

[s] d Ω ≥ −

 Γ

 +

− Q p + Q p · NdΓ + B + + B − d Ω , Ω

which gives [s] ≥ −divΣ Qp + Σ (B) .

(3.38)

Combining relationships (3.37) and (3.38), we get

[Ψ ] + s [T ] + div [T] Δ(Qp ) − [T] Δ(B) ≤ Σ : D(

U+ + U− ) + B p [β ] + H p · grad[β ] − gradT · Σ(Qp ). 2

(3.39)

3.6 The Basic Laws of Mechanics in a Discontinuity Volume

27

Let us note that the right hand side is a scalar product between internal forces and related evolution quantities whereas it is not the case of the left hand side. Let us try to relate [Ψ ] to a scalar product. We have [Ψ ] = Ψ (T + , β + , gradβ + ) − Ψ(T−, β − , gradβ − ) = Ψ (T + , β + , gradβ + ) − Ψ(T+, β − , gradβ − ) + Ψ (T + , β − , gradβ − ) − Ψ(T−, β − , gradβ − ). Because the free energy is a concave function of temperature T , we have

Ψ (T + , β − , gradβ − ) − Ψ(T− , β − , gradβ − ) ≤ −sfe [T] ,

(3.40)

with ˆ T (T − , β − , gradβ − ), −s f e ∈ ∂Ψ ˆ T is the set of the uppergradients of the concave function where ∂Ψ T → ΨT (T, β − , gradβ − ) = Ψ(T, β − , gradβ − ), (see Appendix A). We assume Ψ is a convex function of (β , grad β ). Thus we have

Ψ (T + , β + , gradβ + ) − Ψ(T+ , β − , gradβ − ) ≤ Bfe [β ] + Hfe · grad [β ] , with f e f e B , H ∈ ∂Ψβ ,gradβ (T + , β + , gradβ + ).

(3.41)



The internal forces B f e , H f e depend on the future state (T + , β + , gradβ + ), in agreement with our idea that the constitutive laws sum up what occurs during the discontinuity of the state quantities. It results [Ψ ] + s[T ] + div[T] Δ(Qp ) − [T] Δ(B) ≤ B f e [β ] + H f e · grad[β ] − sfe [T] + s [T] + div[T] Δ(Qp ) − [T] Δ(B) = B f e [β ] + H f e · grad[β ] + (−sfe + s + divΔ(Qp ) − Δ(B)) [T] + Δ(Qp ) · grad[T]. As usual, we assume no dissipation with respect to [T ] and have

Δ (Q p ) = 0, −s f e + s + divΔ(Qp ) − Δ(B) = 0.

28

3 The Basic Laws of Mechanics

Thus

Δ (B)+S = 0,

(3.42)

with S = s f e − s. This relationship splits either the received heat, T + B+ + T − B − , or the received entropy, B + + B − , between the two temperatures, T + and T − . Let us note that this relationship depends on the future state via the average entropy s. ¯ We assume that the interior forces satisfy the inequality   + U + U− + (B p − B f e ) [β ] + (H p − H f e ) · grad[β ] − 2gradT · Qp . 0≤Σ :D 2 (3.43) Remark 3.16. The discontinuity [Ψ ] may be split in a different manner [Ψ ] = Ψ (T + , β + , gradβ + ) − Ψ(T−, β − , gradβ − ) = Ψ (T + , β + , gradβ + ) − Ψ(T−, β + , gradβ + ) + Ψ (T − , β + , gradβ + ) − Ψ(T−, β − , gradβ − ). We get

Ψ (T + , β + , gradβ + ) − Ψ(T− , β + , gradβ + ) ≤ −ˆsfe [T] , with ˆ T (T − , β + , gradβ + ). −sˆf e ∈ ∂Ψ If Ψ it is a convex function of (β , grad β ) ˆ fe · grad [β ] , Ψ (T − , β + , gradβ + ) − Ψ(T− , β − , gradβ − )) ≤ Bˆ fe [β ] + H with f e f e ∈ ∂Ψβ ,gradβ (T − , β − , gradβ − ). Bˆ , Hˆ

The internal forces Bˆ f e , Hˆ f e depend entirely on the past state (T − , β − , gradβ − ). Because we think that the constitutive laws sum up what occurs during the discontinuity, it is compulsory that the internal forces depend on the future state (T + , β + , grad β + ). Thus this splitting of the free energy does not seem as good as the one we have chosen. We may choose a pseudopotential of dissipation    +  U + U− ± ± ± , [β ] , grad [β ] , gradT, E , χ , Φ (Δ E , E , χ ) = Φ D 2

3.6 The Basic Laws of Mechanics in a Discontinuity Volume

29

and the constitutive laws

Σ , (B p − B f e ), (H p − H f e ), −2Q p   +   U + U− , [β ] , grad [β ] , gradT, E± , χ . ∈ ∂Φ D 2

(3.44)

Theorem 3.3. If constitutive laws (3.40), (3.41), (3.42) and (3.44) are satisfied, then the second law is satisfied. Proof. If relationship (3.44) is satisfied, inequality (3.43) is satisfied. Then it is easy to prove that the inequality (3.39) which is equivalent to the second law is satisfied. Remark 3.17. In this chapter, there are discontinuities either with respect to time or with respect to space. It is possible that both discontinuities are concomitant. This is the case when a solid colliding an obstacle is fractured. This non smooth situation is investigated in [114] and [100].

•

Chapter 4

Solid–Liquid Phase Change

Besides the classical ice–water phase change, the solid–liquid phase change exhibits many different behaviours: diffuse with respect to time and space phase change, phase change with voids and bubbles, phase change with different densities, dissipative phase change, irreversible phase change, phase change with thermal memory,... The different predictive theories result from the choices of the free energies and pseudopotentials of dissipation. Examples are given: soil freezing and the insulation of an house in winter, liquid gas underground storage, fossil permafrost.

4.1 The Stefan Problem The classical ice–water phase change or the Stefan problem [2, 34, 96, 149, 200], introduces as state quantities the temperature T and the liquid or water volume fraction β , [3, 133]. The free energy and pseudopotential are [78, 112]

Ψ (T, β ) = −CT ln T − β

L (T − T0) + I(β ), T0

Φ (δ E, χ ) = Φ (gradT, T) =

λ (gradT)2 , 2T

(4.1)

with δ E = gradT and χ = T . Let us comment the different terms of Ψ and Φ : • −CT ln T is responsible for heat storage (C is the heat capacity). • −β (L/T0 )(T − T0 ) is a linear function with respect to β : its derivative being constant with respect to β defines a threshold (we prove below that when (T − T0 ) < 0 the solid phase, the ice, is present and when (T − T0 ) > 0, the liquid phase, the water, is present). The quantity L is the latent heat at phase change temperature T0 .

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 4, © Springer-Verlag Berlin Heidelberg 2012

31

32

4 Solid–Liquid Phase Change

• Indicator function I of the interval [0, 1], (see Appendix A), takes into account the internal constraint on the liquid water volume fraction 0 ≤ β ≤ 1.

(4.2)

• In the classical ice–water phase change, it is assumed that no macroscopic motion occurs. Thus the deformation ε is not a state quantity. We assume that there is no local interaction, thus gradβ does not intervene. • The pseudopotential of dissipation describes the thermal conduction: λ is the thermal conductivity. The constitutive laws are

σ = 0, B∈−

L (T − T0 ) + ∂ I(β ) , H = 0, T0

λ Q = − gradT, T s = C(1 + lnT ) +

L β, T0

where ∂ I is the subdifferential set of function I, (see Appendix A). On the free boundary between ice and water, the constitutive laws are those given in Sect. 3.5. Let us recall that the mass flux through the free boundary is → − − → m = −ρ W where W = W . N is the normal velocity of the free boundary. In the sequel, it is denoted [A] = A2 − A1 , the discontinuity of quantity A on the free boundary (the indices 1 and 2 refer to the two sides of the discontinuity surface). The pseudopotential of dissipation has to imply the well known experimental results: the temperature is continuous on the free boundary and the ice–water phase change is perfectly reversible or non dissipative. Thus we choose the pseudopotential

Φ (m, [T ]) = I0 ([T ]), where I0 is the indicator function of the origin of R (see Appendix A). It results the constitutive laws from (3.36) [T ] = 0, − [Ψ ] =

∂Φ = 0, ∂m

which imply the continuity of the free energy on the free boundary Γ (t). This relationship implies that β (T − T0 ) is continuous. Thus the temperature on the free boundary is T0 because [β ] = 0: the phase change occurs only at this temperature which is, of course, called the phase change temperature.

4.1 The Stefan Problem

33

4.1.1 The Equations They result from the equations of motion, the entropy balance and the constitutive laws. We have assumed that the ice–water system does not move. Thus there is only the equation of microscopic motion. Moreover we assume from now on, that there is no volume exterior source of exterior work A = 0 (in the sequel we are going to assume when needed, no surface exterior source of work a = 0). Outside the free boundary Γ (t) we have

∂ ln T L ∂β + − λ Δ ln T = R , in Ω \Γ (t), ∂t T0 ∂ t L − (T − T0 ) + ∂ I(β )  0 , in Ω \Γ (t) , T0

C

λ

∂ ln T = π , on ∂ Ω , ∂N

T (x, 0) = T 0 (x) , β (x, 0) = β 0 (x) , on Ω ,

(4.3) (4.4) (4.5) (4.6)

where R and π are the volume and surface entropy sources. On free boundary Γ (t), we have the constitutive laws [T ] = [ln T ] = 0 , T = T0 ,

(4.7)

and the entropy balance m [s] − [Q] = 0, or     L ∂ ln T λ ∂T = λ , ρ W [β ] = T0 T ∂N ∂N

(4.8)

or   ∂T , ρ W L [β ] = λ ∂N because T = T0 . This relationship is the classical energy balance on the free boundary of the Stefan problem. The functions T 0 and β 0 are the initial temperature and liquid water content which have to satisfy an obvious compatibility condition. Equation (4.4) results from the equation of microscopic motion. It gives the value of the temperature difference, T − T0 , versus the water volume fraction. It is easy to read this relationship:

34

4 Solid–Liquid Phase Change

1. If T − T0 < 0, then β = 0 and there is ice. 2. If T − T0 > 0, then β = 1 and there is liquid water. 3. If T − T0 = 0, then 0 ≤ β ≤ 1 and there is a mixture of ice and liquid water. Remark 4.1. It is reasonable that the microscopic motions do not intervene on the free boundary because the microscopic equation of motion (4.4) does not involve derivatives with respect to space. The macroscopic equation of motion which contains space derivative does not intervene because we have assumed that the macroscopic velocities U are zero. Remark 4.2. If the term related to β in the free energy −β

L (T − T0 ) + I(β ), T0

is replaced by a quadratic function L T0



 β2 −β (T − T0 ) + (T1 − T0 ) + I(β ), 2

with T1 ≥ T0 , the relationship (4.4) becomes −

L {(T − T0 ) − β (T1 − T0 )} + ∂ I(β )  0 , in Ω . T0

The phase change occurs for T0 ≤ T ≤ T1 with

β=

T − T0 , T1 − T0

and for T > T1 there is liquid: β = 1, for T < T0 there is solid: β = 0. When the temperature is increasing phase change begins at T = T0 and ends at T = T1 . This type of phenomenon occurs in shape memory alloys (see the references on shape memory alloys).

4.1.2 A Way to Solve the Equations Where the Free Boundary Is a By-Product: The Freezing Index Let us note that (4.3) and (4.4) in the distributive sense are valid outside free boundary Γ (t) and on the free boundary where they are equivalent to (4.7) and to (4.8). We may integrate (4.3) or the entropy balance (3.14) with respect to time and get

4.1 The Stefan Problem

35

s(t) − s(0) + −

t 0  t 0

divQdt = Q · Ndt =

 t 0

 t 0

Rdt , in Ω ,

π dt , on ∂ Ω .

It is possible to prove that  t 0

divQdt = div

 t

Qdt. 0

Assuming the Fourier law, it is also possible to prove that  t 0

Qdt = −λ

 t 0

gradln Tdt = −λ grad

 0

t

 ln Tdt .

Thus we have s(t) − s(0) − λ Δ

λ





t

=

ln T dt 0

∂

t

0 ln T dt

∂N

 =

 t 0

 t 0

Rdt , in Ω ,

π dt , on ∂ Ω .

(4.9)

By letting u(x,t) =

 t 0

ln T (x,t)dt,

we get s(x,t) = C(1 + lnT ) +

∂u L L β = C(1 + ) + β . T0 ∂t T0

The equation of motion (4.4) T − T0 ∈ ∂ I(β ), is equivalent to  ∂u − T0 ∈ ∂ I(β ) . exp ∂t 

(4.10)

Equation (4.10) may be solved and the water volume fraction may be eliminated to give

36

4 Solid–Liquid Phase Change

L ∂u ∂u C(1 + ) + ∂ I ∗ (exp( ) − T0 ) − λ Δ u = s(0) + ∂t T0 ∂t

λ

∂u = ∂N

 t 0

 t 0

Rdt , in Ω ,

π dt , on ∂ Ω ,

(4.11)

where I ∗ is the dual function of I (see Appendix A or [112, 172]). This partial differential equation completed by the boundary condition (4.9) and initial condition u(x, 0) = 0 may be solved, [41]. There exists a variational formulation because ln T and u being continuous on the free boundary, gradu involves no surface measure and has only density with respect to the Lebesgue measure. For the numerical methods, it is also possible to use the energy balance instead of the entropy balance. The resulting equation for the freezing index defined by w(x,t) =

 t 0

T (x,t)dt,

is C

∂w ∂w + L∂ I ∗ ( − T0 ) − λ Δ w = e(0) + ∂t ∂t λ

∂w = ∂N

 t 0

 t 0

rdt , in Ω ,

ϖ dt , on ∂ Ω ,

(4.12)

where r = T R and ϖ = T π are assumed to be known functions. The main numerical advantage of (4.11) and (4.12) is that they do not involve the free boundary which is a by-product given by the level set either ∂ u/∂ t = ln T0 or ∂ w/∂ t = T0 , [32, 112]. Approximations useful for numerics are given in [51, 55, 146].

4.2 Examples The computer program CESARGEL, [141], solving (4.12) may be used for engineering purposes. Figure 4.1 shows the temperature inside an insulated house in winter, [89]. Figure 4.2 shows the extension of the permafrost during the last 120,000 years in some place in France [150]. Time in year is the abscissa and depth in meter is the ordinate. Applications to soil freezing are given in [33, 128, 135, 198].

4.3 Unfrozen Water in the Frozen Zone When freezing a water saturated porous medium, experiments show that the water freezes progressively. There exists unfrozen liquid water in the frozen zone, i.e., the zone where T < T0 , [112]. The predictive theory which takes this phenomenon

4.3 Unfrozen Water in the Frozen Zone

37

Fig. 4.1 The temperature in winter of the wall of a warm house and the temperature in the surrounding ground. The temperatures are computed with the freezing index w(x,t). The temperature scale in C is on the right: red color is warm and blue is cold

into account introduces a function h(β ) in the free energy. Its derivative is directly related to the unfrozen water content (see (4.14) below). The free energy and pseudopotential are

Ψ (T, β ) = −CT ln T − β Φ (gradT, T) =

L (T − T0 ) + h(β ), T0

λ (gradT)2 . 2T

Convex function h takes into account the internal constraint (4.2) by being equal to +∞ outside interval [0, 1]. The constitutive laws are

σ = 0, B∈−

L (T − T0) + ∂ h(β ) , H = 0, T0

λ Q = − gradT, T s = C(1 + lnT ) +

L β. T0

Subdifferential set ∂ h(β ) is ⎧ dh ⎪ ⎨ dβ (0) + ∂ I(0), i f β = 0, dh ∂ h(β ) = dβ (β ), i f 0 < β < 1, ⎪ ⎩ dh (1) + ∂ I(1), i f β = 1, dβ with ∂ I(0) = R− and ∂ I(0) = R+ .

(4.13)

38

4 Solid–Liquid Phase Change

Fig. 4.2 The extension of permafrost in the last 120, 000 years at some place in France. The 0◦C curves in the time-depth plane. The results depend on the snow cover on the ground (three scenarii have been investigated). The blue domain is the time-depth domain where the temperature is negative. One may note than during one of the cold periods the soil was frozen up to 300 m

4.3.1 The Equations The equations resulting from the equations of motion, balance of energy and constitutive laws are outside the free boundary Γ (t) L ∂β ∂ ln T + − λ Δ ln T = R , in Ω \Γ (t), ∂t T0 ∂ t L − (T − T0 ) + ∂ h(β )  0 , in Ω \Γ (t), T0

C

(4.14)

4.3 Unfrozen Water in the Frozen Zone

λ

39

∂ ln T = π , on ∂ Ω , ∂N

T (x, 0) = T 0 (x) , β (x, 0) = β 0 (x) , in Ω , and on the free boundary Γ (t) (which exists if h(β ) is constant on some interval) where [Ψ ] = 0, which implies that T − T0 = 0,     L ∂ ln T λ ∂T ρ W [β ] = = λ . T0 T ∂N ∂N Equation (4.14) which results from the equation of microscopic motion, gives the temperature versus the unfrozen water content. Thus function h may be measured with experiments, [112]. Measurement of unfrozen water content versus temperature, [87, 88], gives function (dh/d β )(β ) depending on T (relationship (4.14)).

4.3.2 An Example An example in Fig. 4.3 shows the temperature in the ground surrounding a decommissioned natural gas storage. The CESARGEL computer program, [141], involves

Fig. 4.3 The temperature in the ground around a decommissioned liquefided natural gas storage, 3 years after the removing of the liquefied gas. Part of the soil is still frozen. The soil has heaved due to the cryogenic suction inducing water motion, [33]. The temperature scale in C is given on the left: red color is warm and blue is cold

40

4 Solid–Liquid Phase Change

function h(β ). Moreover, a computation of the frost heave induced by the water motion is shown. Due to the cryogenic suction induced by freezing, a displacement of the ground up to 0.5 m occurs, [33, 112]. One may remark that ground freezing and thawing are slow phenomena, the Fig. 4.3 shows the temperature 11, 340 days, more than 3 years, after the removing of the natural liquefied gas.

4.4 Phase Change with a Freezing Fringe In some phase changes, the phenomena are spread over a small zone around the surface where the temperature is equal to the phase change temperature T0 . This freezing fringe results from spatial interactions. Thus the gradient of β intervene and we choose, [112]

Ψ (T, β , gradβ ) = −CT ln T − β Φ (gradT, T) =

k L (T − T0 ) + I(β ) + (gradβ )2 , T0 2

λ (gradT)2 . 2T

Parameter k measures the intensity of the local interactions: when k is large the freezing fringe is widely spread, when k is small, the freezing fringe is narrow and tends to be a surface in dimension three, a line in dimension two. The term involving gradβ is related to interface properties which influence the phase change phenomenon. For instance, a free energy involving quantity gradβ takes into account surface tension effects. The constitutive laws are

σ = 0, B∈−

L (T − T0 ) + ∂ I(β ) , H = kgradβ , T0

λ Q = − gradT, T s = C(1 + lnT ) +

L β. T0

4.4.1 The Equations The equations of the predictive theory are

∂ ln T L ∂β + − λ Δ ln T = R , in Ω , ∂t T0 ∂ t L −kΔ β − (T − T0 ) + ∂ I(β )  0 , in Ω , T0

C

4.5 Phase Change with Voids and Bubbles

λ

41

∂ ln T ∂β =π , k = 0 , on ∂ Ω , ∂N ∂N

T (x, 0) = T 0 (x) , β (x, 0) = β 0 (x) , in Ω . In this model there is no free boundary. The phase change is spread in the freezing fringe. The indicator function I(β ) may be replaced by a function h(β ).

4.5 Phase Change with Voids and Bubbles Bubbles or voids may appear in phase change. In frozen ice, it may be seen small bubbles. In cast iron voids or bubbles are present. In this situation there are two phase volume fractions β1 and β2 , β = (β1 , β2 ), [124]. The two phases do not fill the whole volume. As for an example, we assume that β1 is the liquid volume fraction, water for instance, and β2 is the solid volume fraction, ice for instance. The voids volume fraction is ν = 1 − β1 − β2 . We assume, for the sake of simplicity, the same constant density ρ (i.e., the material is incompressible) and the same velocity ∂ u/∂ t for the two phases (u is the small displacement). The mass balance is

∂ (β1 + β2 ) + (β1 + β2 ) divU = 0. ∂t

(4.15)

The equations of motion for the microscopic motions responsible of the evolution of β1 and β2 and for the macroscopic motion resulting from the body and surface exterior forces f and g are  −

B1 B2





H1 + div H0



 = 0 , in Ω,

H1 H0

 · N = 0 , on ∂ Ω,

divσ + f = 0 , in Ω , σ · N = g , on ∂ Ω, where works B1 , B2 , work flux vectors H1 , H2 and stress σ are the interior forces. We assume that the evolution is quasistatic and that there is no exterior source of work producing the phase change. The volumic free energy we choose is 2 k Ψ (T, ε , β , gradβ ) = ∑ βj Ψj + IK(β ) + |gradβ |2 , 2 j=1

where Ψj denotes the free energy of phase j , IK is the indicator function (see Appendix A) of the convex set K = {(β1 , β2 ) |0 ≤ β1 ≤ 1, 0 ≤ β2 ≤ 1, β1 + β2 ∈ [0, 1] }.

42

4 Solid–Liquid Phase Change

The term IK (β ) + (k/2) |gradβ |2 is an interaction or mixture free energy. The effect of IK (β ) is to guarantee that the fractions β1 and β2 take admissible physical values. Let us note that even if the free energy of the voids phase is 0, the voids phase has physical properties due to the mixture free energy which depends on the gradients of β1 and of β2 . The gradients are related to the interfaces properties: gradβ1 describes properties of the voids–liquid interface and gradβ2 describes properties of the voids– solid interface, for instance surface tension. In this setting, the voids have a role in the phase change and make it different from a phase change without voids. The model is simple and schematic but it may be upgraded by introducing sophisticated interaction free energy depending on β and on gradβ . Moreover, for the free energies of the phases, we consider the following simplified expressions

Ψ1 (T, ε ) = −CT ln T −

L (T − T0 ) + L (T ), T0

1 Ψ2 (T, ε ) = −CT ln T + {λe (trε )2 + 2μe ε : ε } + L (T ), 2 where λe and μe are the elasticity Lam´e parameters. The linear function L (T ) is defined by L (T ) = CT0 ln T0 + C(1 + lnT0 )(T − T0 ). It has been chosen in such a way that within the small perturbation assumption when T − T0 = θ , with θ small compared to T0 , the quantity −CT ln T + L (T ) is equivalent to its leading term − CT ln T + L (T )

C (T − T0 )2 . 2T0

(4.16)

Remark 4.3. The free energy of a material is defined up to a linear function of the temperature. The free energies Ψ1 and Ψ2 are the free energies of the phases of the same material. Then they are defined up to any linear function of T but, of course, this linear function has to the same for Ψ1 and Ψ2 . Thus the volumic free energy is

Ψ (T, ε , β , gradβ ) = (β1 + β2) (−CT ln T + L (T)) −

k β1 L β2 (T − T0 ) + {λe (trε )2 + 2 μe ε : ε } + IK (β ) + |gradβ |2 . T0 2 2

For the pseudopotential of dissipation with quantity χ = (T, β1 , β2 ), we choose

4.5 Phase Change with Voids and Bubbles

Φ (gradT, ε˙ , =

∂β , T, β1 , β2 ) ∂t

(β1 + β2 )λ c (gradT)2 + 2T 2

+ I0 (

43



∂β ∂t

2 +

β1 {λv (trε˙ )2 + 2μv ε˙ : ε˙ } 2

∂ (β1 + β2) + (β1 + β2 ) divU), ∂t

(4.17)

where ε˙ = ∂ ε /∂ t and I0 is the indicator function of the origin. The viscosity parameters of the fluid are λv and μv . The last term in (4.17) is zero if mass balance (4.15) is satisfied and it is +∞ otherwise. In other words we may say that the presence of the last term in (4.17) is due to the fact that mass balance (4.15) is an internal constraint between velocities and it must be included in the expression of the pseudopotential of dissipation Φ . Let us recall that the pseudopotential accounts for the properties of the velocities. We will see that this term is related to the pressure in the system. Parameter c is the viscosity parameter of the microscopic motion. The constitutive laws are

σ = β1 (λv (trε˙ )1 + 2μvε˙ ) + β2 (λe (trε )1 + 2μeε ) − (β1 + β2 ) p1,       − TL0 (T − T0 ) B1 −CT ln T + L (T ) + ∈ 1 2 −CT ln T + L (T ) B2 2 {λe (trε ) + 2 μe ε : ε }   ∂β 1 +∂ IK (β ) + c −p , ∂t 1 λ (β1 + β 2 ) gradT, T T L s = (β1 + β2)C ln + β1 , T0 T0

Hi = kgradβi , Q = −

where 1 is the identity matrix, p is the pressure defined by − p ∈ ∂ I0

 ∂ (β1 + β2 ) + (β1 + β2 )divU . ∂t

This constitutive law implies that the mass balance is satisfied because the subdifferential set   ∂ (β1 + β2 ) + (β1 + β2 ) divU , ∂ I0 ∂t is not empty. The stress σ is an elastic stress in the solid part and a viscous stress in the fluid.

44

4 Solid–Liquid Phase Change

The small perturbation assumption for T and β1 + β2 are: θ = T − T0 is small compare to T and (β1 + β2) − (β1 (0) + β2(0)) is small compare to (β1 (0) + β2 (0)). For the sake of simplicity, we assume there is no voids at the initial time: β1 (0) + β2 (0) = 1. With the small perturbation assumption, the constitutive laws are 

σ = β1 (λv (trε˙ )1 + 2 μvε˙ ) + β2 (λe (trε )1 + 2μe ε ) − p1,   L    ∂β − T0 θ 1 B1 + ∂ IK (β ) + c ∈ −p , ∂ t B2 1 0 λ Hi = kgradβi , Q = − gradT, T L θ s = C(1 + ) + β1 , T0 T0

where the deformations ε and temperatures θ at power 2 have been neglected in agreement with the small deformation assumption (see Appendix B). Remark 4.4. An other way to have simple constitutive laws is to assume that the voids are filled with vapor with heat capacity C and thermal conductivity λ . The functions Ψ and Φ become

Ψ (T, ε , β , gradβ ) = −CT ln T −

β1 L β2 (T − T0 ) + {λe (trε )2 + 2μe ε : ε } T0 2

k +IK (β ) + |gradβ |2 , 2 and

∂β , T, β1 , β2 ) ∂t   c dβ 2 β1 λ 2 + {λv (trε˙ )2 + 2μv ε˙ : ε˙ } (gradT) + = 2T 2 dt 2 Φ (gradT, ε˙ ,

+I0 (

∂ (β1 + β2 ) + (β1 + β2 ) divU). ∂t

They give the following constitutive laws

σ = β1 (λv (trε˙ )1 + 2μvε˙ ) + β2 (λe (trε )1 + 2 μe ε ) − (β1 + β2 ) p1, ⎛ ⎞ L   − (T − T0 ) ∂β 1 ⎜ ⎟ T0 −p , B ∈ ⎝1 ⎠ + ∂ IK (β ) + c ∂ t 1 2 {λe (trε ) + 2μe ε : ε } 2

4.5 Phase Change with Voids and Bubbles

45

λ Hi = kgradβi , Q = − gradT, T L s = C(1 + lnT ) + β1 . T0

4.5.1 The Equations The equations of the predictive theory are L ∂ β1 λ C ∂θ + − Δ θ = R , in Ω , T0 ∂ t T0 ∂ t T0     ∂β − TL0 θ 1 −p  0 , in Ω , − kΔ β + ∂ IK (β ) + c ∂t 1 0   ∂u ∂u div β1 {λv (trε ( ))1 + 2 μvε ( )} + β2 {λe trε (u)1 + 2μeε (u)} ∂t ∂t −gradp + f = 0 , in Ω,

∂ ∂u (β1 + β2 ) + div = 0 , in Ω, ∂t ∂t ∂θ ∂β = T0 π , k = 0 , σ N = g , on ∂ Ω , λ ∂N ∂N θ (x, 0) = θ 0 (x) , β (x, 0) = β 0 (x) , u(x, 0) = u0 (x) , in Ω .

(4.18)

Let us note that the pressure intervene in both equations for macroscopic and microscopic motions which are coupled. Mathematical results may be found in [124].

4.5.2 The Case Where There Is Only Liquid: The Cavitation Phenomenon Let us assume that there are only water and voids, 0 < β1 < 1, β2 = 0 and interpret equation (4.18) assuming an homogeneous state. We have 

∂ IK (β1 , 0) = with Q ≥ 0. Equation (4.18) is

0 −Q

 ,

46

4 Solid–Liquid Phase Change

 c

∂ β1 ∂t



 +

0

0 −Q



 +

− TL0 θ 0

 −p

  1  0. 1

It results p = −Q < 0,

∂ β1 L = p + θ. ∂t T0

c

The pressure is negative in the water. If p + (L/T0 )θ < 0,

∂ β1 < 0. ∂t The voids volume fraction increases and from the mass balance we get divU > 0. This is the cavitation phenomenon which results from the pressure becoming negative. Let us also note that if p+

L θ ≥ 0 , θ ≥ 0, T0

we may have only liquid at an equilibrium: β1 = 1 and (∂ β1 /∂ t) = 0. Thus large temperature and large pressure make the liquid to be the only phase present at an equilibrium.

4.5.3 The Case There Is Only Solid: Soil Mechanics In case there is only solid and voids, β1 = 0, 0 < β2 < 1, (4.18) becomes  c

0

∂ β2 ∂t



 +

−P 0



 +

− TL0 θ 0

 −p

  1  0. 1

We get that the voids volume fraction either increases or decreases if the pressure is either low or large. This is a property of soils mechanics, [162]. Let us also note that if p ≥ 0 , θ ≤ 0,

4.5 Phase Change with Voids and Bubbles

47

we may have only solid at an equilibrium: β2 = 1 and (∂ β2 /∂ t) = 0. Thus non negative pressure and non positive temperature make the solid to be the only phase present at an equilibrium. We conclude that the pressure p is the main practical tool to govern the voids volume fraction and that the temperature T is the main practical tool to govern the phase change.

4.5.4 Numerical Examples We investigate the evolution of a one-dimensional continuous body occupying domain (−h/2, h/2). The numerical results are due to Francesco Ascione, [12]. We denote w(z,t), the vertical displacement; θ (z,t) = T (z,t) − T0 , the difference between the absolute temperature and phase change temperature (one may think of the Celsius temperature); βi (z,t) the phase volume fractions, with i = 1 for liquid phase volume fraction and i = 2 for solid phase volume fraction; p(z,t), the pressure. The exterior forces applied to the structure are F, the vertical body force and G, the surface traction on the top of the structure; R(z,t), is the volume entropy source and πtop = π (z = h/2,t), πbottom = π (z = −h/2,t) are the surface entropy flows. Quantities R(z,t)T0 and π (z = ±h/2,t)T0 are the volume and surface heat sources. We assume the solid is fixed on its bottom face. We investigate the influence of the physical parameters which intervene in the phase change phenomenon: in example 1, the influence of the external load when cavitation occurs; in examples 2 and 3, the effect of microscopic motion parameter k which accounts for spatial interactions; in examples 4 and 5, the influence of the other microscopic motion parameter, c which accounts for the dissipative effects; finally, in example 6, the influence of pressure p. For all the examples we assume the following values of the boundary conditions: at top , z =

h ∂θ ∂ β1 ∂ β2 ,λ = T0 πtop , k =0, k = 0 , σzz = G, 2 ∂z ∂z ∂z

h ∂θ ∂ β1 ∂ β2 = T0 πbottom , k =0, k = 0 , w = 0. at bottom , z = − , −λ 2 ∂z ∂z ∂z The initial conditions are for all the examples

θ (z, 0) = 0 , β1 (z, 0) = 0 , β2 (z, 0) = 1 , w(z, 0) = 0, except in the example 1 (cavitation) where the volume fraction β1 and β2 have inverted values. The thickness is h = 1 m, the surface entropy fluxes are πtop = 0.5 W/(m2 K), πbottom = −0.05 W/(m2 K), phase change temperature is T0 = 273.15 K, the other mechanical and thermal parameters have usual values, elasticity parameters λe = 10.4 109 Pa, μe = 3.9 109 Pa, viscosity parameters

48

4 Solid–Liquid Phase Change

λv = 1.00 109 Pa s, μv = 1.827 109 Pa s, latent heat 3.35 108 J/m3 , heat capacity 4.186 106 J/ (kg K) and thermal conductivity 2.23 W/ (K m). The top tension G, phase change viscosity c and local influence parameter k change in the different examples. Note that the thermal action mostly heats the structure which is frozen at initial time (except for the cavitation experiment where it is liquid water).

4.5.5 Example 1: The Cavitation Phenomenon We investigate the cavitation phenomenon: an increase of voids volume fraction and a consequent decrease of the water volume fraction due to a low pressure resulting from a tension applied at the top of the structure. The parameters are, respectively: G = 1.01 105 Pa (G is a tension), k = 107 J/m, c = 107 (J s)/m3 . The numerical results are reported in Fig. 4.4a–e where it is shown the evolution of θ , v, β1 , w and p. As expected solid volume fraction remains null and the voids volume fraction increases due to the low pressure. The voids and water volume fractions at the two ends of the structure at final time are: at bottom z = −h/2, v = 0.278283, β1 = 0.721362, and at top z = h/2, v = 0.411954, β1 = 0.588632.

4.5.6 Example 2: The Non Dissipative Spatial Interaction Is Important To emphasize the effect of the interaction parameter k, we choose an extreme large value for this parameter, almost equal to 1020 J/m. The other two variable data are c = 107 (Js)/m3 and G = −1.01 105 Pa. Volume fractions β1 and β2 are shown in Fig. 4.5a, b. Functions z → βi (z) become constant. Note that at the both ends of the structure the two volume fractions have the same values: at bottom z = −h/2, β1 = 0.003632 and β2 = 0.9963 and at top z = h/2, β1 = 0.003728 and β2 = 0.996296. A large spatial interaction parameter makes the evolution of the βi ’s very dependent of the neighbourhood. On the contrary a very low interaction parameter makes the evolution decoupled and independent of the neighbourhood, as shown in the following section.

4.5.7 Example 3: No Spatial Interaction In opposition to previous example 2, parameter k is now almost null, k = 10−20 J m. The other data are those of the previous example. It results that the evolution of the

4.5 Phase Change with Voids and Bubbles

a

49

b

t 800000 600000 400000 200000

t 800000 600000 400000 200000

20

0.4

10

0.3 0.2

0

0.1 0

–10

0 z

0.2 0.4

–0.2

–0.4

0.4

d c

0 z

0.2

–0.2

–0.4

800000 t 600000 400000 200000

t 800000 600000 400000 200000

0.0002 0.0001

1 0.9

0.00005

0.8

0

0.7 0.6

0 z

0.2 0.4

–0.2

0.2

0.4

–0.4

0 z

–0.2

–0.4

t 800000 600000 400000 200000

e

0 –5x106 –1x107

0.2 0.4

0 z

–0.2

–0.4

Fig. 4.4 Example 1. The cavitation phenomenon is due to low pressure p shown in (e), resulting in creation of voids (voids volume fraction is shown in (b)). Temperature θ , water volume fraction β1 , displacement w are shown in (a), (c) and (d). Note that the vertical scale of figures giving βi versus z and t are not the same in the figures of this section

βi ’s are independent of their neighbourhood. Because temperature θ governing the evolution of the βi ’s, is smooth with respect to space, the βi ’s are also smooth with respect to space. Water and ice volume fractions are shown in Fig. 4.6a, b. At the two ends of the structure, the volume fractions have independent values: at bottom z = −h/2, β1 = 0.278283 and β2 = 0.721362, and at top z = h/2, β1 = 0.411954 and β2 = 0.588632.

50

4 Solid–Liquid Phase Change

a

b

t

8000 6000

t

8000 6000

4000

4000

2000

2000 0.003

1

0.002

0.999 0.998

0.001

0.997 –0.5

0 –0.5 –0.25

0 z

0.25 0.5

0 z

0.25 0.5

–0.25

Fig. 4.5 Example 2. In case there is an important spatial interaction in the phase change, the whole structure is instantaneously affected by the boundary actions. Water volume fraction β1 and ice volume fraction β2 are shown in (a) and (b)

a

b

t 800000 600000 400000 200000 0.4 0.3 0.2

800000 t 600000 400000 200000 1 0.9 0.8

b1

0.4

0.2

0 z

–0.2

b2

0.7 0.6

0.1 0 –0.4 0.4

0.2

0 z

–0.2

–0.4

Fig. 4.6 Example 3. In case there is no spatial interaction in the phase change, the whole structure is not instantaneously affected by the boundary actions. Water volume fraction β1 and ice volume fraction β2 are shown in (a) and (b)

4.5.8 Example 4: Dissipative Phase Change In these next two sections we investigate the effect of the dissipative microscopic motion parameter c. It controls the velocity of the phase change. For this purpose, we assume firstly it has a very large value, c = 1020 J s/m3 . The other two variable data, k and G have the following values, respectively: k = 107 J/m and G = −1.01 105 Pa. Volume fractions β1 and β2 are shown in Fig. 4.7a, b. As expected, functions t → βi (t) are constant. For instance, we have: at bottom z = −h/2, β1 = 0.00154885, β2 = 0.998433, and at top z = h/2, β1 = 0.00711434, β2 = 0.992941.

4.5.9 Example 5: Non Dissipative Phase Change Contrary to the previous case, we choose c to be almost null with value, c = 10−20 (J s)/m3 , and keep the values of k and G as in example 4. In agreement with

4.5 Phase Change with Voids and Bubbles

a

8000

51

b

t 6000 4000

t

8000 6000

4000 2000

2000 0.006

1

0.004

0.998

0.002

0.996

0 –0.5 0.25 0.5

0 z

0.994 –0.5

–0.25 0.25 0.5

–0.25

0 z

Fig. 4.7 Example 4. A dissipative phase change is slow. Water volume fraction β1 and ice volume fraction β2 are shown in (a) and (b)

a

b

t 800000 600000 400000 200000

t 800000 600000 400000 200000 1

0.4 0.3

0.2 0.4

0 z

–0.2

0.9

0.2

0.8

0.1 0

0.7 0.6

–0.4 0.2 0.4

0 z

–0.2

–0.4

Fig. 4.8 Example 5. Phase change is rapid in case it is not dissipative. Water volume fraction β1 and ice volume fraction β2 are shown in (a) and (b)

the predictive theory, phase change is almost instantaneous as seen in Fig. 4.8a, b. The corresponding volume fractions β1 and β2 are: at bottom z = −h/2, β1 = 0.278283, β2 = 0.721362, and at top z = h/2, β1 = 0.411954, β2 = 0.588632. As c decreases, the velocity of the phase change increases.

4.5.10 Example 6: Effect of the Pressure In this last example, we investigate the effect of the pressure p on the phase change phenomenon. We assume parameters k and c have the following values, respectively: 107 J/m and 107 (J s)/m3 . We apply a very large compression at the top of the structure with value G = −1.01 105 Pa. In agreement with the predictive theory, very high pressure prevents ice to be transformed into water even if temperatures are positive, as shown in Fig. 4.9a, b. The volume fractions have the following values: at bottom z = −h/2, β1 = 0.001500, β2 = 0.998433, and at top z = h/2, β1 = 0.007114, β2 = 0.992941.

52

4 Solid–Liquid Phase Change

a

b

t

8000 6000

t

8000 6000

4000

4000

2000

2000 0.006

1

0.004

0.998

0.002

0.996

0

0.994 –0.5

–0.5 0.25 0.5

–0.25

0 z

0.25 0.5

0 z

–0.25

Fig. 4.9 Example 6. High pressure prevents phase change. Water volume fraction β1 and ice volume fraction β2 are shown in (a) and (b)

Let us note that the pressure has a paramount importance whereas it does not intervene when voids are not taken into account. When the pressure is low cavitation occurs and when it is large it prevents the phase change. The predictive theory which spares the number of parameters to be measured with experiments, has a scope large enough to account for the basic physical features of phase changes involving voids. Of course there are many possible upgrading sophistications in order to give a finer prediction of some aspects of the phenomena. Remark 4.5. In case we assume the densities of the liquid and solid to be constant but slightly different, the mass balance becomes with ρ 0 = ρ1 β1 (0) + ρ2β2 (0)

∂ (ρ1 β1 + ρ2 β2 ) + ρ 0 divU = 0. ∂t In case we have no voids and a mixture of ice and water, β1 + β2 = 1, at an equilibrium, we have 

∂ IK (β1 , β2 ) =

P P

 ,

with P ≥ 0. It results that 

P P



 +

− TL0 θ 0



 −p

ρ1 ρ2

  0,

or −

L θ − p(ρ1 − ρ2 ) = 0, p ≥ 0. T0

(4.19)

In this situation the phase change temperature depends on the pressure. It is known that this is actually the case for the ice–water phase change. Because ρ1 > ρ2 the phase change temperature decreases when the pressure increases. A consequence of

4.6 Solid–Liquid Phase Change with Different Densities

53

this property is that the ice melts at the bottom of a glacier and lubricates the rock ice contact surface allowing the downhill motion of the glacier. The previous relationship shows that the pressure is 0, when the phase change occurs at temperature 0◦ C. Thus the stress −ρ 0 p1 resulting from the mass balance constraint is the pressure with respect to the atmospheric pressure. This phenomenon is investigated in the following section where the macroscopic equation of motion is not as sophisticated as it is in this section.

4.6 Solid–Liquid Phase Change with Different Densities Let us consider water in a rigid and impermeable container and freeze the container. Because density of ice ρ2 is lower than density of water ρ1 it seems impossible that water transforms into ice: the water remains liquid even at low temperature. Experiments, for instance the freezing of a glass bottle filled with water, show that the water pressure increases up to the rupture of the bottle. When the container is not impermeable, freezing may produce a non homogeneous material, for instance water ice or sorbet. We investigate this problem, [125]. The theory takes into account the density difference between ice and water but the equation for the macroscopic motion is less sophisticated than that of the previous section: we assume the ice has zero velocity. Water velocity is U1 .

4.6.1 The Mass Balance It is

ρ1

dβ 1 ∂ β2 + ρ2 + ρ1 β1 divU1 = 0 , in Ω , dt ∂t −ρ1 β1 U1 · N = ϖ , on ∂ Ω ,

(4.20)

where ϖ is the liquid water intake and N is the outward normal vector to domain Ω occupied by the ice water mixture.

4.6.2 The Equations of Motion For the macroscopic motion of the water, it is

ρ1

d(β1 U1 ) = divσ1 + f1 , in Ω, dt σ1 N = g1 , on ∂ Ω ,

(4.21)

54

4 Solid–Liquid Phase Change

where σ1 is the water stress, f1 is the action of the exterior on the liquid water, for instance the friction of the water on the solid ice phase and g1 is the boundary exterior action on the liquid water. For the microscopic motions responsible for the ice water phase change, it is  −

B1 B2





 H1 = 0 , in Ω , H2   H1 · N = 0 , on ∂ Ω , H2

+ div

(4.22)

assuming no exterior source of work.

4.6.3 The Entropy Balance For the sake of simplicity, till the end of this section we assume the small perturbation assumption. In this setting the entropy balance is

∂s + divQ = R ∂t  1 ∂ β1 ∂ β2 σ1d : D(U1 ) + Bd1 + + Bd2 T ∂t ∂t +Hd1 · grad

 ∂ β1 ∂ β2 + Hd2 · grad − gradT.Q , in Ω , ∂t ∂t

where the internal forces are split into dissipative forces indexed by dissipative interior forces indexed by nd .

4.6.4 The Constitutive Laws The non dissipative forces are defined with the water free energy

σ1nd = 0, Bnd 1 =

∂Ψ ∂Ψ , Hnd , 1 = ∂ β1 ∂ gradβ1

Bnd 2 =

∂Ψ ∂Ψ , Hnd . 2 = ∂ β2 ∂ gradβ2

(4.23) d

and non

4.6 Solid–Liquid Phase Change with Different Densities

55

Free energy Ψ takes into account the internal constraint on the state quantities, for instance (β1 , β2 ) ∈ K = {(γ1 , γ2 ) |0 ≤ γ1 ≤ 1, 0 ≤ γ2 ≤ 1, γ1 + γ2 ∈ [0, 1] }. We choose

Ψ (θ , β1 , β2 , gradβ1 , gradβ2 ) = −β1 −

C1 2 C2 2 θ − β2 θ 2T0 2T0

k k β1 L θ + IK (β1 , β2 ) + (gradβ1 )2 + (gradβ2 )2 , T0 2 2

where θ = T − T0 is the Celsius temperature assumed to be small. Thus we have within the small perturbation assumption C1 C2 β1 L θ + β2 θ + , T0 T0 T0  nd   L  − T0 θ B1 + ∂ IK (β1 , β2 ), = Bnd 0 2  nd    H1 gradβ1 = k . Hnd gradβ2 2 s = β1

(4.24)

(4.25)

The dissipative forces are defined with pseudopotential of dissipation Φ which takes into account the internal constraints related to the velocities. The mass balance is such a constraint. For instance, we may choose

∂ β1 ∂ β2 ∂ β1 ∂ β2 , , grad , grad , D(U1 ), gradθ ) ∂t ∂t ∂t ∂t ∂ β1 ∂ β2 + ρ2 + ρ1 β1 divU1 ) = I0 (ρ1 ∂t ∂t     λ c ∂ β 1 2 c ∂ β2 2 + + + (gradθ )2 , 2 ∂t 2 ∂t 2T0

Φ(

where D(U1 ) is the strain rate. Let us denote −p an element of the subdifferential set ∂ I0 . Thus the constitutive laws are

σ1d = −pρ1β1 1 , ∂ β1 − pρ1 , Hd1 = 0, ∂t ∂ β2 − pρ2 , Hd2 = 0. Bd2 = c ∂t Bd1 = c

Quantity pρ1 β1 is the pressure in the liquid phase.

(4.26) (4.27) (4.28)

56

4 Solid–Liquid Phase Change

4.6.5 Approximation of the Equation for the Macroscopic Motions The equation of motion (4.21) and constitutive law (4.26) give

ρ1

∂ (β1 U1 ) = −grad(pρ1 β1 ) + f1 . ∂t

We assume the volume exterior force f1 results mainly from friction on the solid ice phase and that this force is proportional to the relative velocity f1 = −

(ρ1 β1 )2 U1 . m

Because the force has to be 0 when there is no water, we assume force f1 to be proportional to the square of the liquid water density. We neglect the acceleration forces and have (ρ1 β1 )2 U1 = −grad(pρ1 β1 ), m which is within the small perturbation theory

ρ1 β1 U1 = −mgradp.

(4.29)

Parameter m/ρ1 is the mobility of the water and velocity β1 U1 is the filtration velocity which is measured in experiments. This relationship is in agreement with the Darcy law. Remark 4.6. Force f1 is the density of force applied to the liquid water with respect to volume d Ω . The actual density of force applied to the volume occupied by the liquid water β1 d Ω is 2 ˜f1 = f1 = − ρ1 β1 U1 . β1 m

Force ˜f1 is proportional to β1 as expected. It is also proportional to the water momentum (ρ1 β1 ) U1 .

4.6.6 The Equations for the Phase Change They result from mass balance (4.20), equation of motion (4.21), (4.22), entropy balance (4.23) and constitutive laws (4.24), (4.27), (4.25), (4.28) and Darcy law (4.29)

4.6 Solid–Liquid Phase Change with Different Densities

57

∂ β1 ∂ β2 + ρ2 − mΔ p = 0 , in Ω , ρ1 ∂t ∂t        L  ∂ β1 − T0 θ Δ β ρ1 1 −p + + ∂ IK (β1 , β2 )  0 , in Ω , c ∂∂βt2 − k Δ β ρ 0 2 2 ∂t ∂θ ∂ β1 ∂θ ∂ β2 ∂ β1 + θ 0C1 + β20C2 + θ 0C2 +L ∂t ∂t ∂t ∂t ∂t −λ Δ θ = T0 R , in Ω . β10C1

(4.30)

This set of equation is completed by boundary conditions on ∂ Ω

∂p + α p (p − pext ) = 0, ∂N ∂ β1 ∂ β1 =0, = 0, ∂N ∂N ∂θ λ + αθ (θ − θext ) = 0, ∂N m

where α p and αθ measure the hydraulic and thermal conductivity of the container boundary (α p = 0 if the boundary is watertight, α p = ∞ if the boundary is connected to a water supply with pressure pext ), and initial conditions

β1 (x, 0) = β10 (x) , β2 (x, 0) = β20 (x) , θ (x, 0) = θ 0 (x) . This set of partial differential equations allows to compute pressure p(x,t), liquid water and ice volume fractions β1 (x,t), β2 (x,t) and temperature θ (x,t) depending on the external actions resulting from the exterior pressure pext (x,t) and exterior temperature θext (x,t) and rate of heat production T0 R(x,t) very often equal to 0. In an engineering situation the governing action is the exterior temperature cooling and heating the container. Mathematical results are reported in [125]. Remark 4.7. More sophisticated boundary condition result from more sophisticated physical boundary properties. The boundary of the container can be semipermeable, in this case we have

ρ1 β1 U1 · N = −α p (p − pext )− , or m

∂p − α p (p − pext )− = 0. ∂N

This boundary condition means that when the pressure is lower than the exterior pressure pext , water flows inside the container but when the pressure is larger than the exterior pressure, no water flows outside .

58

4 Solid–Liquid Phase Change

The pressure may also be controlled on the boundary m

∂p + ∂ I− (p − pext )  0. ∂N

Water flows outside the container in order to maintain the pressure lower than the outside pressure pext which may be the atmospheric pressure. Remark 4.8. The energy balance, the last (4.30), may be simplified within the small perturbation assumption (reference temperature θ 0 is small) (β10C1 + β20C2 )

∂θ ∂ β1 +L − λ Δ θ = T0 R , in Ω . ∂t ∂t

4.6.7 An Example: Liquid Water in an Impermeable Container We have

β10 = 1 , β20 = 0 , θ 0 > 0. The boundary condition for the pressure is

∂p = 0. ∂N We investigate what occurs when the container is cooled. In order to look for closed form solutions, we assume all the quantities are homogeneous. Assuming the temperature is known, we focus on (4.30) which describes the phase change. Because pressure is homogeneous, (4.30) is  ∂β  1

c

∂t ∂ β2 ∂t



ρ1 −p ρ2



 +

− TL0 θ 0

 + ∂ IK (β1 , β2 )  0.

When the temperature is positive the solution of this equation is β1 = 1 and β2 = 0. When the container is cooled and its temperature becomes negative, the phase change which is expected does not occur. Indeed according to the mass balance, to values β1 = 1 and β2 = 0 and condition β1 + β2 ≤ 1, we should have

ρ2

∂ β2 ∂ β1 ∂ β1 ∂ β2 ∂ β1 + ρ1 =0, 0, ∂t ∂t ρ2 ∂ t

4.6 An Other Example: Ice in an Impermeable Container

59

which contradicts the third condition: thus phase change is impossible. Let us check that the water remains liquid (∂ β1 /∂ t = 0) even if temperature is negative. We must have    L  − T0 θ ρ1 + ∂ IK (1, 0)  0, + −p ρ2 0 with 

∂ IK (1, 0) =

P Q

 , P ≥ 0 , Q ≤ P,

or pρ1 +

L θ ≥ 0, T0 0 ≤ p(ρ1 − ρ2 ) +

L θ. T0

(4.31)

These two conditions may be satisfied because ρ1 − ρ2 > 0. The minimum value of the pressure is p=−

L θ. (ρ1 − ρ2 )T0

Thus when cooling the container, water remains liquid and pressure increases. This is in agreement with experiment: a glass container filled of water explodes when freezing. The explosion occurs when the pressure overcomes the glass resistance.

4.6.8 An Other Example: Ice in an Impermeable Container We have

β10 = 0 , β20 = 1 , θ 0 < 0, and we investigate what occurs when the container is heated. In order to look for closed form solutions we assume all the quantities are homogeneous. Equation (4.30) which describes the phase change becomes  ∂β  1

c

∂t ∂ β2 ∂t



ρ1 −p ρ2



 +

− TL0 θ 0

 + ∂ IK (β1 , β2 )  0.

(4.32)

60

4 Solid–Liquid Phase Change

When temperature is negative, β1 = 0, β2 = 1, is solution with p = 0, for instance 

− TL0 θ 0

 + ∂ IK (0, 1)  0.

When temperature becomes positive, β1 = 0, β2 = 1, is no longer solution because 

∂ IK (0, 1) =

P Q

 , Q ≥ 0 , P ≤ Q,

implying 

ρ1 −p ρ2



 +

− TL0 θ ) 0



 +

P Q

  0,

or Q = pρ2 ≥ 0, Q − P = p(ρ2 − ρ1 ) −

L θ ≥ 0. T0

(4.33)

The two last inequalities are impossible. Thus the water cannot remain frozen. It thaws according to equation  ∂β  1

c

∂t ∂ β2 ∂t



ρ1 −p ρ2



 +

− TL θ 0 0

 = 0,

with

ρ2

∂ β2 ∂ β1 + ρ1 = 0, ∂t ∂t

due to the mass balance. The pressure is p=−

ρ1 L θ. 2 (ρ1 + ρ22 )T0

The solution satisfies

ρ22 L ∂ β1 ∂ β2 − ρ2 ρ1 L θ >0, θ < 0, = = 2 2 ∂t ∂ t c(ρ1 + ρ2 )T0 c(ρ12 + ρ22 )T0 2  ρ2 − ρ2 ρ1 L ∂ β 2 ∂ β1 + = θ < 0, ∂t ∂t c(ρ12 + ρ22 )T0

4.6 An Other Example: Ice in an Impermeable Container

61

as long as β1 ≤ 1, β2 ≥ 0. The evolution stops when β2 is 0. The final value of β1

β1 =

ρ2 < 1, ρ1

shows that there are voids in agreement with the homogeneity assumption. We have    L  ρ2 − T0 θ ρ1 + ∂ IK ( , 0)  0, + −p ρ1 ρ2 0 with

∂ IK (

ρ2 , 0) = ρ1



0 Q

 , Q ≤ 0,

giving the pressure p=−

L θ. ρ1 T0

4.6.9 An Other Example: Freezing of a Water Emulsion in an Impermeable Container We have 0 < β10 < 1 , β20 = 0 , θ 0 > 0. The emulsion, a mixture of voids and water, is cooled. Assuming an homogeneous evolution, it may be seen that the water freezes till either β1 = 0, (in this case β10 < ρ2 /ρ1 and the cooling results in a mixture of ice and voids), or β1 + β2 = 1, (in this case β10 ≥ ρ2 /ρ1 and the cooling results in a mixture of ice and liquid water, β1 = (ρ1 β10 − ρ2)/(ρ1 − ρ2 ), β2 = (ρ1 − ρ1 β10 )/(ρ1 − ρ2 )). This mixture is a water ice or sorbet.

4.6.10 The Non Homogeneous Evolution In practice cooling and heating are performed with surface actions. Thus the temperature is non homogeneous. One may expect that the resulting mixture is non homogeneous. In the last example, the ice volume fraction is larger in the neighbourhood of the surface of the container than in the middle where the liquid volume fraction is more important. The evolution is given by solving the set of partial differential equations.

62

4 Solid–Liquid Phase Change

4.6.11 The Motion of Glaciers In case we have an homogeneous mixture of ice and water without voids at an equilibrium,

β1 + β2 = 1 ,

∂ β 1 ∂ β2 = = 0, ∂t ∂t

(4.32) gives 

ρ1 −p ρ2



 +

− TL0 θ 0

 +

  P = 0, P

with P ≥ 0. We get from this equation − p(ρ1 − ρ2 ) −

L θ , p ≥ 0. T0

(4.34)

This is relationship (4.19) up to the pressure unit, either p with pascal unit, or ρ p with p with meter unit for relationship (4.34). Because ρ1 > ρ2 the phase change temperature decreases when the pressure increases. A consequence of this property is that the ice melts at the bottom of a glacier and lubricates the rock ice contact surface allowing the downhill motion of the glacier.

4.7 Dissipative and Irreversible Phase Change When the phase change is progressive or when the microscopic motion is viscous, there is dissipation with respect to the phase volume fraction, [112, 127]. The evolution can even be irreversible (for instance when cooking an egg, [127]). The Ψ and Φ with parameter χ = T are k L (T − T0 ) + I(β ) + (gradβ )2 , T0 2   dβ c dβ 2 dβ λ 2 Φ (gradT, + I−( ), , T) = (gradT) + dt 2T 2 dt dt

Ψ (T, β , gradβ ) = −CT ln T − β

where I− is the indicator function of the set of the non positive numbers. It takes into account the irreversibility of the phase change dβ ≤ 0. dt

4.8 The Phase Volume Fraction Depends on the Temperature

63

The constitutive laws are

σ = 0, B∈−

L dβ dβ + ∂ I(β ) + ∂ I−( ) , H = kgradβ , (T − T0 ) + c T0 dt dt

λ Q = − gradT, T s = C(1 + lnT ) +

L β. T0

4.7.1 The Equations The equations of the predictive theory, assuming small perturbations, are C

∂ ln T L ∂β + − λ Δ ln T = R , in Ω , ∂t T0 ∂ t

c

L ∂β ∂β − kΔ β − (T − T0 ) + ∂ I(β ) + ∂ I−( )  0 , in Ω , ∂t T0 ∂t

λ

∂ ln T ∂β = π, k = 0 , on ∂ Ω , ∂N ∂N

T (x, 0) = T 0 (x) , β (x, 0) = β 0 (x) , in Ω . When the evolution is not irreversible, the subdifferential ∂ I− (∂ β /∂ t) is absent from the equations. The phase change occurs at temperatures which are close to T0 as it may be seen in equation c

∂β L − kΔ β − (T − T0 ) + ∂ I(β )  0. ∂t T0

Other results on dissipative phase change and related problems may be found in [14, 35, 47, 48, 56, 68, 70, 73, 127, 144, 160, 207, 208].

4.8 The Phase Volume Fraction Depends on the Temperature It is possible that volume fractions depend on the temperature. To account for this phenomenon, we choose a the pseudopotential of dissipation Φ depending on χ = (T, β ) k¯ k L (T − T0 ) + I(β ) + β 2 + (gradβ )2 , T0 2 2      dβ  dβ c dβ 2 λ (gradT)2 + Φ (gradT, , T, β ) = + K(T, β )   , dt 2T 2 dt dt

Ψ (T, β , gradβ ) = −CT ln T − β

64

4 Solid–Liquid Phase Change

where K(T, β ) is a non negative function. The physical parameter k¯ is positive. In this situation, the pseudopotential of dissipation depends on the parameters T, β of E. The constitutive laws are

σ =0, B∈−

L dβ dβ + ∂ I(β ) + K(T, β )sgn( ) , H = kgradβ , (T − T0 ) + k¯ β + c T0 dt dt

where the sign graph is defined by ⎧ ⎪ ⎨ 1, sign(x) = [−1, 1], ⎪ ⎩ −1,

if x > 0, if x = 0, if x < 0.

4.8.1 The Equations The equations of the predictive theory, assuming small perturbations, are C

∂ ln T L ∂β + − λ Δ ln T = R , in Ω , ∂t T0 ∂ t

c

L dβ ∂β − kΔ β − (T − T0) + k¯ β + ∂ I(β ) + K(T, β )sgn( )  0 , in Ω , ∂t T0 dt

λ

∂ ln T ∂β =π , k = 0 , on ∂ Ω , ∂N ∂N

T (x, 0) = T 0 (x) , β (x, 0) = β 0 (x) , in Ω . It is possible to prove that the volume fraction β is within an interval depending on the temperature. For instance, if c∂ β /∂ t − kΔ β = 0 and 0 < β < 1, we have 1 k¯



   L 1 L (T − T0 ) − K(T, β ) ≤ β ≤ ¯ (T − T0 ) + K(T, β ) . T0 k T0

Thanks to a convenient pseudopotential of dissipation, the predictive theory accounts for the experimental fact: the volume fraction depends on the temperature. Let us stress the large possibilities offered by the theory: there are so many possible pseudopotentials of dissipation! Mathematical results may be found in [13, 14, 72].

4.9 Phase Change with Thermal Memory

65

4.9 Phase Change with Thermal Memory The heat flux vector may depend on the past, [40]. We define the new state quantity G(t) =

1 2

 +∞ 0

h(s)p(s,t) · p(s,t)ds,

with p(s,t) =

 t (t−s)

gradT(τ ) dτ .

This function depends on the history of the temperature gradient. Properties of kernel h may be found in [40]: h : (0, +∞) → R is continuous, positive and decreasing. We assume the free energy depends on G. Thus it depends on the past. We may have

Ψ (T, β , gradβ , G) = −CT ln T − β

k L (T − T0) + I(β ) + (gradβ )2 + G. T0 2

The pseudopotential of dissipation with χ = T is chosen as

Φ (gradT,

λ dβ c , T) = (gradT)2 + dt 2T 2



dβ dt

2 .

The constitutive laws result from upgraded laws of thermodynamics introducing non dissipative entropy flux vector, [40]

σ = 0, B∈−

L ∂β (T − T0 ) + ∂ I(β ) + c , H = kgradβ , T0 ∂t

λ Q = − gradT − T

 t

−∞

k(t − s)gradT(s) ds,

L β, T0

s = C(1 + ln T ) +

where function k is defined by k (s) = −h(s) and k(+∞) = 0. The heat flux vector depends on the past, on the history of the temperature gradient gradT. This model is due to Pierluigi Colli and to Elena Bonetti.

4.9.1 The Equations We assume that the history term  0

div

−∞

k(t − s)gradT(s) ds,

66

4 Solid–Liquid Phase Change

is known and include it in the generalized entropy sources, still denoted by R and π . Thus the equations of the predictive theory, assuming small perturbations, are C

∂ ln T L ∂β + − λ Δ ln T − k ∗ Δ T = R , in Ω , ∂t T0 ∂ t

c

L ∂β − kΔ β − (T − T0 ) + ∂ I(β )  0 , in Ω , ∂t T0

λ

∂ ln T ∂ (k ∗ T ) ∂β + =π , k = 0 , on ∂ Ω , ∂N ∂N ∂N

T (x, 0) = T 0 (x) , β (x, 0) = β 0 (x) , in Ω , where ∗ denotes the usual time convolution product over the interval (0,t), i.e., for a and b (a ∗ b)(t) :=

 t 0

a(t − s)b(s) ds.

Mathematical results may be found in [40, 71]. Other models and memory effects may be found in [91–94].

Chapter 5

Shape Memory Alloys

5.1 Introduction Shape memory alloys are mixtures of many martensites and of austenite. The composition of the mixture varies: the martensites and the austenite transform into one another. These phase changes can be produced either by thermal actions or by mechanical actions. The striking properties of shape memory alloys result from interactions between mechanical and thermal actions [29, 136, 183]. We describe a macroscopic predictive theory which can be used for engineering purposes. The phase volume fractions, which are state quantities, are subjected to constraints (for instance their sum is one when there are no voids and no interpenetration in the mixture). It is shown that most of the properties of shape memory alloys result from a careful treatment of those internal constraints, [109, 110, 112, 119]. Other microscopic and macroscopic models, applications and results may be found in [1, 19, 20, 23, 30, 95, 155, 157, 164, 169, 170, 173, 178, 190, 191, 202].

5.2 The State Quantities We deal only with macroscopic phenomena and macroscopic quantities. To describe the deformations of the alloy, both the macroscopic small deformation ε and the temperature T are chosen as state quantities. The properties of shape memory alloys result from martensite–austenite phase changes produced either by thermal actions (as it is usual) or by mechanical actions. On the macroscopic level, quantities are needed to take those phase changes into account. For this purpose, the volume fractions βi of the martensites and austenite are chosen as state quantities. We think that this choice is the simplest which can be made. Again to be simple, we assume that only two martensites exist together with austenite. The volume fractions of the martensites are β1 and β2 . The volume M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 5, © Springer-Verlag Berlin Heidelberg 2012

67

68

5 Shape Memory Alloys

fraction of austenite is β3 . We assume there are local microscopic interactions responsible for the twinning and choose the gradβi as state quantities. The volume fractions are not independent: they satisfy constraints, called internal constraints 0 ≤ βi ≤ 1,

(5.1)

due to the definition of volume fractions. We assume that no void can appear in the mixture, then β1 + β2 + β3 ≥ 1, and that no interpenetration of the phases can occur, then β1 + β2 + β3 ≤ 1. Thus the βi ’s satisfy an other internal constraint

β1 + β2 + β 3 = 1 .

(5.2)

These internal constraints are physical properties due to their definitions or to their mechanical properties. The macroscopic state quantities are E = (ε , β1 , β2 , β3 , gradβ1 , gradβ2 , gradβ3 , T) .

(5.3)

The case where voids can appear is investigated in Sect. 5.6.

5.3 The Equations of Motion They are for the macroscopic motions

ρ

dU = divσ + f, in Ω, σ N = g, on ∂ Ω, dt

(5.4)

where ρ is the density, dU/dt the acceleration, σ the stresses in the alloy which occupies domain Ω , with boundary ∂ Ω and outward normal vector N. The alloy is loaded by macroscopic body forces f and by surface tractions g.

5.3.1 The Equations for the Microscopic Motions When Voids and Interpenetration Are Possible For the microscopic motions, they result from the principle of virtual power ∀γ j , j = 1, 2, 3, 3

−∑

 

j=1 Ω

3  3   H j · gradγ j + Bj γ j d Ω + ∑ A j γ jdΩ + ∑ j=1 Ω

j=1 ∂ Ω

a j γ j d Γ = 0,

5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible

69

where the γ j are virtual velocities, the H j and B j are the internal forces and the A j and a j are the external forces. These equations of motion are divHj − Bj + Aj = 0, in Ω, Hj · N = aj , on ∂ Ω, j = 1, 2, 3.

(5.5)

Let us note that the virtual velocities do not satisfy

γ1 + γ2 + γ3 = 0,

(5.6)

as the actual velocities. The satisfaction of this relationship by the actual velocities will result from the constitutive laws which will introduce reactions to this internal constraint. The set of the virtual velocities is a very large one. This point of view is useful in case voids can appear

β1 + β2 + β3 ≤ 1, as seen in Sect. 4.5 or in case interpenetration can occur

β1 + β2 + β 3 ≥ 1 .

5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible An other way to proceed is to choose starting from the beginning that the internal constraint (5.2) is satisfied and to have virtual velocities which satisfy relationship (5.6). The principle gives ∀γi , i = 1, 2, 2

−∑



i=1 Ω 2

+∑



i=1 Ω

{(Hi − H3 ) · gradγi + (Bi − B3 )γi } d Ω 2

(Ai − A3 )γi d Ω + ∑



i=1 ∂ Ω

(ai − a3 )γ j d Γ = 0 .

By letting Hi = Hi − H3 , Bi = Bi − B3 , Ai = Ai − A3 , ai = ai − a3 , the equations of motion become divHi − Bi + Ai = 0, in Ω, Hi · N = ai , on ∂ Ω, i = 1, 2 . In the sequel, we assume no exterior work Ai and ai .

(5.7)

70

5 Shape Memory Alloys

The way relationship (5.2) is used implies that β3 is no longer a state quantity. The set of states quantities is E = (ε , β1 , β2 , gradβ1 , gradβ2 , T) . We denote β the vector (βi , i = 1, 2), B the vector (Bi , i = 1, 2) and H the matrix (Hi , i = 1, 2). Remark 5.1. The second set of (5.7) may be recovered from the first one (5.5) when voids and no interpenetration are impossible by eliminating the reactions due to internal constraint (5.2).

5.4.1 The Entropy Balance By denoting s the entropy, Q the entropy flux vector, the entropy balance is ds 1 + divQ = R + dt T

  dβ d d dβ d , in Ω , + H : grad σ : D(U) + B · dt dt

−Q · N = π , on ∂ Ω ,

(5.8) (5.9)

where R is the exterior volume rate of entropy that is supplied to the alloy, π is the rate of entropy that is supplied by contact action, D(U) is the strain rate. The dissipative internal forces are σ d , Bd , H d . Let us recall that in this setting the second law is 

T > 0,  dβ dβ ≥ 0. + H d : grad σ d : D(U) + Bd · dt dt

5.4.2 The Free Energy As already said, a shape memory alloy is at the macroscopic level a mixture of three martensite and austenite phases with volume fractions βi . The volumic free energy of the mixture we choose is

Ψ =

3

∑ βiΨi + T h(E) + IC(β ),

i=1

= β1Ψ1 (E) + β2Ψ2 (E) + (1 − β1 − β2 )Ψ3 (E) + T h(E) + IC (β ) , (5.10)

5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible

71

where the Ψi ’s are the volume free energies of the i phases and T h is a free energy describing interactions between the different phases. We have said that internal constraints are physical properties. Being physical properties, we decide to take them into account with the two functions we have to describe the material, i.e., the free energy Ψ and the pseudopotential of dissipation Φ . The pseudopotential describes the kinematic properties, i.e., properties which depend on the velocities. The free energy describes the state properties. Obviously the internal constraints (5.1) and (5.2) are not kinematic properties. Thus we take them into account with the free energy Ψ which involves the quantity IC (β ) where IC is the indicator function (see Appendix A) of the convex set, see Fig. 5.1 C = {(γ1 , γ2 ) ∈ R2 ; 0 ≤ γi ≤ 1; γ1 + γ2 ≤ 1} . The simplest choice we can make for h(β ) is h(β ) = 0. There is no interaction, besides constraints (5.1) and (5.2), between the different phases in the mixture. For the volume free energies, we choose 1 ε : K1 : ε + σ1 (T ) : ε − C1 T ln T, 2 1 Ψ2 (E) = ε : K2 : ε + σ2 (T ) : ε − C2 T ln T, 2 1 L Ψ3 (E) = ε : K3 : ε − (T − T0 ) − C3 T ln T, 2 T0

Ψ1 (E) =

where Ki are the volume elastic tensors and Ci the volume heat capacities of the phases. The stresses σ1 and σ2 depend on temperature T . The quantity L is the latent heat martensite-austenite volume phase change at temperature T0 (see below).

Fig. 5.1 Vector −Bnd (E) for different deformations ε at medium temperature

72

5 Shape Memory Alloys

Remark 5.2. To make the model more realistic, we can introduce two temperatures to characterize the transformation: T0 , the temperature at the beginning of the phase change and T f the temperature at the end. The interaction free energy is completed by h(β ) = (l2 /T0 )(T0 − T f )(β3 )2 , [22, 112, 156, 185]. Because we want to describe the main basic properties of the shape memory alloys, we assume that the elastic matrices Ki and the heat capacities Ci are the same for all of the phases ∀i = 1, 2, 3, Ci = C, Ki = K . Always for the sake of simplicity, we assume that

σ1 (T ) = −σ2 (T ) = −τ (T ) . Concerning the stress τ (T ), it is known that at high temperature the alloy has a classical elastic behaviour. Thus τ (T ) = 0 at high temperature, and we choose the schematic simple expression

τ (T ) = (T − Tc )τ , for T ≤ Tc , τ (T ) = 0, for T ≥ Tc ,

(5.11)

with τ 11 ≤ 0 and assume the temperature Tc is larger than T0 . With those assumptions, it results 1 L Ψ (E) = ε : K : ε − (β1 − β2 )τ (T ) : ε − (1 − β1 − β2 ) (T − T0 ) − CT ln T . 2 T0 Remark 5.3. In a simple setting, we may choose 

τ (T ) = α (T )1 = (T − Tc )τ =

0, i f T ≥ Tc , (T − Tc )τ¯11 1, i f T ≤ Tc ,

with τ¯11 < 0. Matrix 1 is the identity matrix. Function α (T ) is shown in Fig. 5.2.

Fig. 5.2 Function α (T ). It is 0 at high temperature. The different characteristic temperatures are in the following order: Te < T0 < Tc

5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible

73

Remark 5.4. It is known that elasticity tensor K1 does depend on temperature T , [132]. Thus the entropy of martensite is s1 = −

∂Ψ1 ∂1 σ1 1 ∂ K1 =− ε : :ε− : ε + C1 (1 + lnT ) . ∂T 2 ∂T ∂T

Because free energy has to be a concave function of temperature, −

∂12 σ1 1 ∂ 2 K1 C1 ∂ 2Ψ1 ∂ s1 = − = ε : : ε − :ε+ , ∂T2 ∂T 2 ∂T2 ∂T2 T

has to be non negative. Let us assume, in agreement with experiments, [132], that −(∂ 2 K1 /∂ T 2 ) is positive definite, i.e., T → ε : K1 (T ) : ε is a concave function. Thus we have −

∂ 2Ψ1 ∂12 σ1 ∂ 2 K1 ∂12 σ1 C1 . ≥ : : + ∂T2 ∂T2 ∂T2 ∂T2 T

It results that if C1 ≥ −

T 2



∂12 σ1 ∂ 2 K1 ∂12 σ1 : : ∂T2 ∂T2 ∂T2

 ,

free energy is a concave function of T . This relationship holds in the neighbourhood of Tc in case function α (T ) is replaced by a smooth approximation. Previous inequality is easily satisfied if the variation of α (T ) from 0 to τ¯11 is spread on a sufficiently large interval of temperature (i.e., if the curvature ∂12 σ1 /∂ T 2 is not too large in a neighbourhood of Tc and null outside). In case s1 is discontinuous at temperature Tc as in relationship (5.11) [Ψ ] = 0 . We may choose in agreement with constitutive laws (3.36) [T ] = 0 . These properties are important to solve the entropy balance. But let us note that within the small perturbation assumption entropy is s1 = −

∂Ψ1 = C1 (1 + lnT ), ∂T

a concave function of T . Thus remaining within this assumption, we assume elasticity tensors do not depend on T .

74

5 Shape Memory Alloys

5.4.3 The Constitutive Laws The entropy is s=−

∂Ψ d τ (T ) L = (β1 − β2 ) : ε + (1 − β1 − β2 ) + C(1 + lnT ), ∂T dT T0

or s=−

∂Ψ L = (1 − β1 − β2 ) + C(1 + lnT ), ∂T T0

within the small perturbation assumption. Remark 5.5. With the assumptions of Remark 5.4, we have dτ ∂Ψ 1 ∂K =− ε : : ε + (β1 − β2 ) : ε + C(1 + lnT ) ∂T 2 ∂T dT L +(1 − β1 − β2 ) , T0

s=−

which is an increasing function of T because 1 ∂ 2 K1 d2τ C − ε: : ε + (β1 − β2 ) 2 : ε + 2 2 ∂T dT T C d2τ 1 ∂ 2 K1 : ε − 2 : ε + ≥ 0, ≥− ε : 2 2 ∂T dT T where σ1 (T ) = −σ2 (T ) = −τ (T ), Ki (T ) = K(T ) and Ci = C, i = 1, 3. The non dissipative forces σ nd , Bnd depending on E and non dissipative reaction Bndr depending on (E, x,t) are

σ nd (E) =

∂Ψ (E) = K : ε − (β1 − β2)τ (T ), ∂ε

Bnd (E) =

L L ∂Ψ (E) = {−τ (T ) : ε + (T − T0 ), τ (T ) : ε + (T − T0 )}, ∂β T0 T0

(5.12)

(5.13) H nd (E) =

∂Ψ (E) = kgradβ , ∂ (gradβ )

Bndr (E, x,t) ∈ ∂ IC (β (x,t)) .

(5.14) (5.15)

Relationships (5.12)–(5.15) are the state laws. From the preceding formulas, the smooth part of the free energy is differentiable and the non-smooth part is subdif-

5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible

75

ferentiable, i.e., the subdifferential set ∂ IC (β ) is not empty (see Appendix A). The quantity Bndr (E, x,t) is the thermodynamical reaction to internal constraints (5.1) and (5.2). It is related to β by state law (5.15). This implies that the subdifferential set ∂ IC (β ) is not empty, thus that β ∈ C, which means that internal constraints (5.1) and (5.2) are satisfied. State law (5.15) has an other meaning besides implying that the internal constraints are satisfied: it gives the value of the reactions to these internal constraints. The dissipative forces are defined via a pseudopotential of dissipation Φ . From experiments, it is known that the behaviour of shape memory alloys depends on time, i.e., the behaviour is dissipative. We define pseudopotential of dissipation with χ =T





∂ β k ∂ β 2 λ ∂β



+ (gradT)2 ,

, gradT, T) = c

+ Φ( ∂t ∂t 2 ∂t

2T where the euclidean norm is



  

∂β

∂ β1 2 ∂ β2 2

=

+ ,

∂t

∂t ∂t with c ≥ 0 and k ≥ 0. Of course this expression is induced by experimental results: the first term is related to the permanent deformations exhibited by experiments; the second term is related to the viscous aspect. It has a smoothing effect. The dissipative forces are

∂ Φ (δ E, χ ) = σ d , Bd , Qd ,

(5.16)

where ∂ Φ is the subdifferential of the convex function Φ with respect to 

δE =

 ∂ε ∂β , , gradT , ∂t ∂t

(5.17)

see the explicit computation of ∂ Φ below. Relationship (5.16) gives the constitutive laws

σ = σ d (δ E, χ ) + σ nd (E),

(5.18)

B = Bd (δ E, χ ) + Bnd (E) + Bndr (E, x,t),

(5.19)

H = H nd (E),

(5.20)

−Q = Qd (δ E, χ ),

(5.21)

where E = (ε , β , gradβ , T). The functions σ d , Bd , and Qd are the dissipative or irreversible forces. It can be proved that our choice is such that the internal constraints and the second law of thermodynamics are satisfied [112, 119].

76

5 Shape Memory Alloys

5.4.4 An Example of Non Dissipative Evolution We assume that there is no dissipation and that the material is homogeneous, i.e., gradβ = 0. This assumption is not very realistic but it is a step toward the complete understanding of the constitutive laws. The results from this non dissipative theory are very schematic and crude, but they give some insight into the important role of the reaction Bndr . Let us consider a unidimensional experiment and assume that ε11 is the only non null deformation. Let us focus on the stress σ11 as a function of ε11 when the temperature is fixed. From relationship (5.18) with σ d = 0 and (5.12) it results

σ11 = K1111 ε11 + τ11 (T )(β2 − β1 ),

(5.22)

and from relationships (5.19), (5.13) and (5.15) −

∂Ψ (E) = −Bnd (E) = Bndr ∈ ∂ IC (β ), ∂β

(5.23)

with Bnd (E) = {−τ11 (T )ε11 +

L L (T − T0), τ11 (T )ε11 + (T − T0 )} . T0 T0

The relationship (5.23) means that vector −Bnd (E) is normal to triangle C at the point β r (see Appendix A). Let us consider an experiment at fixed medium temperature (T0 < T < Tc ). We denote ε = ε11 and σ = σ11 . The two components of vector Bnd (E) are shown in Fig. 5.3. When ε = 0, the two components of −Bnd (E) are equal and non positive. Vector −Bnd (E) can be normal to triangle C only at vertex 0 (Fig. 5.1). Thus β3 = 1 and β1 = β2 = 0, there is only austenite. Relationship (5.22) gives σ = 0 for ε = 0. When ε = 0, vector −Bnd (E) is normal to C at vertex 0 if its two

Fig. 5.3 The components of Bnd (E) at medium temperature

5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible

77

components are non positive, i.e., for −

L(T − T0 ) L(T − T0 ) ≤ε ≤ = ε1 . T0 τ11 (T ) T0 τ11 (T )

Thus for ε ∈] − ε1 , ε1 [, there is only austenite, β3 = 1, and from (5.22), σ = K ε (Fig. 5.4). nd For ε = ε1 , Bnd 1 (E) = 0. Vector −B (E) is normal to triangle C on the side OA (Fig. 5.1) and β1 + β3 = 1: there is a mixture of austenite and martensite number one. The stress σ can take any value of the segment [K ε1 , K ε1 − τ11 (T )] (Fig. 5.4). nd nd For ε > ε1 , then −Bnd 1 (E) > 0 and −B2 (E) < 0. Vector −B (E) is normal to triangle C at vertex A (Fig. 5.1) and β1 = 1, there is only the martensite number one and the stress is σ = K ε − τ11 (T ). The increase of deformation produces the martensite–austenite phase change. The same result is obtained when decreasing the deformation: the phase change from austenite to martensite occurs at ε = −ε1 . When ε < −ε1 , there is only martensite number two, β2 = 1, and the stress is σ = K ε + τ11 (T ), (Fig. 5.4). The resulting constitutive law shown on Fig. 5.4 is typical of the behaviour of shape memory alloys at medium temperature.

5.4.5 Latent Heat of Austenite–Martensite Phase Change Let us assume the temperature to be fixed and compute T ds = −T d(

∂Ψ T T ) = −(τ : ε + L )d β2 + (τ : ε − L )d β1 − (β2 − β1 )τ : d ε , ∂T T0 T0

Fig. 5.4 The constitutive law at medium temperature

78

5 Shape Memory Alloys

where s is the entropy. When a phase change occurs at a fixed medium temperature from austenite to martensite number one at the deformation ε = ε1 , i.e., when going from ε slightly lower than ε1 to ε slightly greater than ε1 , the reversible heat received by the material is T Δ s = −L

T + τ 11 ε1 , T0

because Δ β2 = 0, Δ β1 = 1. It has been assumed τ 11 < 0, thus T Δ s < 0. The austenite–martensite phase change is exothermic at medium temperature: when the material is deformed heat is produced. This result is in accordance with experiments [136, 183]. Note that the quantity L is the latent heat of austenite–martensite volume phase change at temperature T0 of the undeformed material (ε1 = 0 and T = T0 ). Within the small perturbation assumption, the previous relationships become T ds = −T d(

T T ∂Ψ ) = −L (d β2 + d β1) = L d β3 , ∂T T0 T0

where β3 = 1 − β1 − β2 is the austenite volume fraction. Note that the internal energy is e = Ψ + T s = CT + Lβ3 , within the small perturbation assumption with L the austenite–martensite phase change latent heat.

5.4.6 The Dissipative Constitutive Laws The dissipative forces are

σ d (δ E, χ ) = 0, Bd (δ E, χ ) ∈ ∂ Φ (

∂β ), ∂t

where the subdifferential set of the non-smooth function (∂ β /∂ t) → Φ is ∂β

∂β ∂β ∂β ∂t

) = c

∂ Φ(

∂ β + k ∂ t , if ∂ t = 0, ∂t



∂t

  ∂β ∂ Φ (0) = S = S ∈ R2 ; S ≤ c, , if =0. ∂t Then the constitutive laws are

σ = σ nd (E) = K : ε − (β1 − β2 )τ (T ),

(5.24)

5.4 Shape Memory Alloys When Voids and Interpenetration Are Impossible

Bnd (E) = {−τ (T ) : ε +

79

L L (T − T0 ), τ (T ) : ε + (T − T0 )}, T0 T0

Bndr (E, x,t) ∈ ∂ I(β (x,t)), B = Bnd (E) + Bndr (E, x,t) + Bd (

∂β ), ∂t

(5.25)

H = kgradβ , −Q = Qd (δ E, T ) =

λ gradT . T

(5.26)

5.4.7 Evolution of a Shape Memory Alloy The equations of the predictive theory for the small displacement u(x,t), the martensites volume fractions β (x,t) and the temperature T (x,t) of a structure Ω made of shape memory alloy are: • The equation for the macroscopic motion div(K : ε (u) − (β1 − β2 )τ (T)) + f = 0, in Ω, σ N = g, on ∂ Ω .

(5.27)

• The equations for the microscopic motions 

−τ (T ) : ε (u) + TL0 (T − T0 ) τ (T ) : ε (u) + TL0 (T − T0 )  ∂β  k ∂ N1 = 0, on ∂ Ω . k ∂∂βN2



 + ∂Φ

∂β ∂t



 + ∂ I(β ) −

kΔ β1 kΔ β 2

 0, in Ω ,

(5.28)

• The entropy balance   d τ (T ) L (β1 − β2) : ε (u) + (1 − β1 − β2 ) + C(1 + lnT ) dT T0   1 ∂β ∂β −λ Δ ln T = · ∂ Φ( ) , in Ω , T ∂t ∂t

∂ ∂t

λ ∂ ln T = π , on ∂ Ω , ∂N

(5.29)

where π is the surfacic rate of entropy provided to the shape memory structure.

80

5 Shape Memory Alloys

These equations are completed by initial conditions for temperature T and volume fractions β T (x, 0) = T0 (x), β (x, 0) = β 0 (x) . It is to be emphasized that the three partial differential equations are coupled in agreement with the fact that thermal actions can produce significant deformations. The equations, mainly the entropy balance, may be simplified within the small perturbation assumption (see Remarks 5.4 and 5.5). For instance, we may choose C

  ∂T ∂ β1 ∂ β2 − λ Δ T = 0, +L + ∂t ∂t ∂t

as an approximation of the entropy balance or of the energy balance. To illustrate the behaviour of shape memory alloys, let us consider a isotherm and homogeneous sample, i.e., gradβ = 0. Equations of motion (5.28) result in non-smooth differential equations. The problem to investigate is the evolution of a material subjected to time dependent exterior actions. In this section an unidimensional experiment is investigated. We choose to apply a deformation ε (t) because it is easier to investigate the structure of the equations and exhibit the hysteretic properties of shape memory alloys. We consider an experiment at medium temperature (T0 < T < Tc ): the applied deformation increases from zero then decreases till the stress is zero. The initial mixture is made of the two martensites with equal volume fractions β1 (0) = β2 (0) = 1/2. The point (σ (t), ε (t)) in Fig. 5.5 follows the path 1, 2, 3, 4, 5, in agreement with experiments. In actual experiments, the temperature evolves due to the heat supplied by the phase changes. The results, for instance, in tension experiments, depend on the way the heat is exchanged with the outside of the sample under consideration. Thus the characterization experiments must take into account both the mechanical and the thermal phenomena. The dissipative phenomena are not as important as thought, [64, 65, 159, 185]. To be didactic, the previous example has been assumed, as already said, to be isothermal.

5.5 Education of Shape Memory Alloys Shape memory alloys can exhibit one shape memory effects or two shape memory effects. A shape memory alloy with one shape memory effect remembers only one shape: the material can be permanently deformed at low temperature by an external mechanical action, then submitted to heating it goes back to its initial form but under cooling it does not recover its deformed shape (Fig. 5.6).

5.5 Education of Shape Memory Alloys

81

Fig. 5.5 The stress σ (t) versus deformation ε (t) in an experiment at medium temperature. In the grey domains the alloy can be in equilibrium

Fig. 5.6 The mechanical behaviour of a non educated shape memory alloy at low temperature in bold line and at high temperature in standard line. A deformed shape memory alloys heated goes back to the initial configuration but cooling of an hot shape memory alloy does not give it the deformed configuration. It is said the non educated shape memory remembers the undeformed configuration

A shape memory alloy with two shape memory effect remembers two shapes: one at low temperature, an other one at high temperature. Following the temperature evolution, it goes from one shape to the other without external action (Fig. 5.7).

82

5 Shape Memory Alloys

Fig. 5.7 The mechanical behaviour of an educated shape memory alloy at high temperature in standard line and at low temperature in bold line. The shape of the material changes reversibly following the temperature. It goes from the reference configuration at high temperature to the deformed configuration at low temperature. It is said the educated shape memory alloys remembers the two configurations

Fig. 5.8 The educated shape memory triangle Ce . The possible compositions of the educated shape memory alloy are within triangle Ce . We assume that the point β 0r = (1/2, 1/2) does not belong to Ce

5.5.1 The Two-Shape Memory Effect The two shape memory effect can be obtained by a special thermomechanical treatment of the alloy which then can remember two shapes, one at low temperature, another one at high temperature. Depending on the temperature the shape goes from one to the other. This treatment is called the education [110, 183]. Its effect is to make one martensite dominant. It is much more present than the other one. From our point of view, the effect of education is to replace triangle C by the flattened triangle Ce (Fig. 5.8): the possible compositions of an educated shape memory alloy are within triangle Ce (Fig. 5.8).

5.5 Education of Shape Memory Alloys

83

Fig. 5.9 Constitutive law of a non dissipative, educated, shape memory alloy at low temperature (the bold line) and at high temperature (the dotted line). When temperature goes back and forth between T + and T − , the alloy goes back and forth between states E + and E − . It remembers the two states!

It is obvious that martensite number one is present to a greater degree than martensite number two. Let us consider an unloaded (σ = 0), educated shape memory alloy at high temperature T + . Its deformation is ε = 0. Let us cool it. The only unloaded (σ = 0) equilibrium state at low temperature T − is (σ = 0, ε = ε2 ). If the temperature is again increased the alloy goes back to the state (σ = 0, ε = 0) and so on (Fig. 5.9). The alloy remembers two shapes: the state E + (ε = 0, σ = 0, β3 = 1, T + ) and the state E − (ε = ε2 , σ = 0, β1 = 1, T − )! One at high temperature and an other one at low temperature. The heat which is received from the exterior during the phase change, for instance when going from E − to E + , is  1  − + − − 2 − e+ r − er = C(T − T ) − K ε2 (T ) − Tc τ 11 ε2 (T ) + L, 2 with ε2 = τ11 (T )/K = (T − Tc )τ 11 /K, (K = K1111 ). Then − + − − 2 2 e+ r − er = C(T − T ) + L − {(T ) − Tc }

τ 211 . 2K

As we expect, it is positive because T − ≤ Tc . Remark 5.6. The one shape memory effect describes what occurs when a non educated alloy is submitted to the same exterior action. The effect is not so striking because it requires mechanical action to go back and forth. In the sequel it is given a macroscopic predictive theory of this education whose basic features have been given in [110].

84

5 Shape Memory Alloys

5.5.2 A Predictive Theory of the Education of Shape Memory Alloys At the macroscopic level the state quantities describing the state of the material are E = (ε , β1 , β2 , gradβ1 , gradβ2 , T). We have to distinguish two materials which have the same state E but have different education. As it has been said the education result in different sets C of R3 of the physically possible or feasible mixtures (β1 ,β2 ), [183]. In order to be simple, we decide to quantify the education by an education index δ which has values between 0 and 1. The non educated material has index 0 and a completely educated material has index 1. Thus the state of a material is defined by (ε , β1 , β2 , β3 , gradβ1 , gradβ2 , δ , T). The feasible mixtures (β1 , β2 ) with education δ belong to the set C(δ ). An educable shape memory alloy is a mixture which can have feasible mixtures (β1 , β2 ) which are in the set C(0) (see Fig. 5.10) C(0) = {(γ1 , γ2 ) |0 ≤ γi ≤ 1; γ1 + γ2 = 1 }, when it is non educated and in C(1) (see Figs. 5.11 or 5.12) C(1) = {(γ1 , γ2 ) |0 ≤ γ1 ≤ 1; 0 ≤ γ2 ≤ βmax ; γ1 + γ2 = 1 },

Fig. 5.10 The triangle C(0) of the feasible mixtures of a non educated material

Fig. 5.11 The possible mixtures of martensites of an educated shape memory alloy are in the triangle C(1). The martensite 1 is much more present than the martensite 2 whose maximum volume fraction is βmax

5.5 Education of Shape Memory Alloys

85

Fig. 5.12 The triangle C(1) of the feasible mixtures of an educated material. The triangle C(δ ) in grey for the education index δ . The projection X of vector Z on the plane (β1 ,β2 ) in standard line for ε small and negative and the projection Y of the same vector in dotted line for ε smaller (larger in absolute value) and negative

when it is completely educated and in the set C(δ ) when the education is uncompleted C(δ ) = {(γ1 , γ2 ) |0 ≤ γ1 ≤ 1; 0 ≤ γ2 ≤ βs (δ ); γ1 + γ2 = 1 }, with

δ=

1 − βs , 1 − βmax

where βs is the ordinate of the vertex M(βs ) of the triangle C(δ ) containing the physically feasible mixtures (β1 ,β2 ), where βmax is the value of βs for a completely educated material (δ = 1), (Figs. 5.11 and 5.12). The volume fraction βmax is the maximum volume fraction of austenite 2 which can be present in an educated shape memory alloy. The properties of the material are defined by its free energy Ψ (E) and by its pseudopotential of dissipation Φ , assumed for the sake of simplicity to depend only on ∂ δ /∂ t = δ˙ , grad T and χ = T

Φ = Φ(

∂δ , gradT, T) . ∂t

The resulting constitutive laws are (see [112] and Sect. 3.3.1)

86

5 Shape Memory Alloys

σ =

∂Ψ ∂Ψ ∂Ψ , , Bi = , Hi = ∂ε ∂ βi ∂ (gradβi )

∂Ψ ∂ Φ + , ∂δ ∂ δ˙ ∂Φ , Q=− ∂ (gradT) 0=

where σ is the stress tensor and Q the entropy flux vector. It is chosen

Ψ (ε , β1 , β2 , gradβ1 , gradβ2 , δ , T) 1 = −Cc T ln T + ε : K : ε − (β1 − β2 )α (T )trε 2 k L Le + (gradβ )2 + (T − T0 )(β1 + β2) − (T − Te )δ + IP (β1 , β2 , δ ), 2 T0 Te where Cc is the heat capacity, K the elasticity tensor, L the austenite–martensite phase change latent heat at temperature T0 , Le the education latent heat at temperature Te (it is assumed Te < T0 < Tc ). The thermal expansion function α (T ) is shown in Fig. 5.2. To make the model more realistic, we can introduce two temperatures to characterize the austenite–martensite phase change: T0 , the temperature at the beginning of the transformation and T f the temperature at the end. The free energy is completed by h(β ) = (L2 /T0 )(T0 − T f )(β3 )2 , [22,155,156,185], (see Remark 4.2 of Chap. 4). The function IP is the indicator function of the pyramid P (Fig. 5.13) of all the feasible austenite–martensite mixtures P=



C(δ )

0≤δ ≤1

= {(γ1 , γ2 ) |0 ≤ γi ≤ 1; γ1 + γ2 = 1; γ2 ≤

1 − δ (1 − βmax) γ1 } . δ (1 − βmax)

Even if the pyramid P, Fig. 5.13, is not a convex set, generalized derivatives of its indicator function can be defined [181]: they are normal vectors to the pyramid. Their properties are sufficient to ensure the mechanical coherence, for instance to ensure that the second principle of thermodynamics is satisfied. The chosen pseudopotential of dissipation is

Φ(

a ∂δ ∂δ λ , gradT, T) = (gradT)2 + ( )2 , ∂t 2T 2 ∂t

5.5 Education of Shape Memory Alloys

87

Fig. 5.13 The pyramid P. The education follows the path AB

where λ , the thermal conductivity, and a, the parameter characterizing the dissipation of the education, are positive. It results the constitutive laws [112]

σ = K : ε + (β2 − β1 )α (T )1,

(5.30)

where 1 is the unit tensor; ⎞ ⎛ ⎞ ⎞ ⎛ −α (T )trε + TL0 (T − T0 ) B1 0 ⎟ ⎜ ⎝ B2 ⎠ ∈ ⎝ α (T )trε + L (T − T0 ) ⎠ + ⎝ 0 ⎠ + ∂ IP(β1 , β2 , δ ), (5.31) T0 a ∂∂δt 0 − LTe (T − Te ) ⎛

e

and Hi = kgradβi ,

λ Q = − gradT . T 5.5.2.1 The Education in Case There Is No Local Microscopic Interaction For the sake of exhibiting the basic education features, we assume that there is no local interactions, i.e. k = 0. It results the equation of motion for the microscopic motions are   B1 =0. B2

88

5 Shape Memory Alloys

The education is a thermomechanical treatment which makes δ to evolve from 0 to 1. To investigate this education, let us consider a non educated material at fixed low temperature T , Te < T < T0 , with the initial composition (β1 , β2 , δ ) = (0, 1, 0) to which is applied the uniaxial deformation ε11 = ε < 0, (εi j = 0 f or (i, j) = (1, 1)). The constitutive law (5.31) and equation of motion give with trε = ε ⎞ ⎛ ⎛ ⎞ ⎛ ⎞ −α (T )ε + TL0 (T − T0 ) 0 0 ⎜ ⎟ ⎝ 0 ⎠ ∈ ⎝ α (T )ε + L (T − T0 ) ⎠ + ⎝ 0 ⎠ + ∂ IP (β1 , β2 , δ ) . T0 0 a ∂∂δt − LTe (T − Te )

(5.32)

e

Thus vector ⎞ −α (T )ε + TL0 (T − T0 ) ⎟ ⎜ Z = − ⎝ α (T )ε + TL0 (T − T0 ) ⎠ , − LTee (T − Te ) + a ∂∂δt ⎛

is normal to the pyramid at the point (β1 ,β2 ,δ ). For the chosen value of ε , vector  X=−

−α (T )ε + TL0 (T − T0 ) α (T )ε + TL0 (T − T0 )

 ,

projection of vector Z on the plane (β1 , β2 ) is shown in a standard line in Fig. 5.12. One can check that vector Z is normal to the pyramid on the segment (A, B) (vector X is normal to triangle C(δ ) at point M(βs )). It results a

Le ∂δ = (T − Te ) + 2(1 − βmax)α (T )ε > 0, ∂t Te

for ε small enough (in absolute value). The education index increases, the education is achieved when δ equals 1. The point representing the mixture has followed the path (A, B), (Fig. 5.13). When ε is a little bit smaller (larger in absolute value) the point representing the mixture (β1 , β2 , δ ) moves at the beginning on the segment AB. But when the point C is reached (Fig. 5.14), vector Z whose projection, Y, on the plane (β1 ,β2 ) is shown in a dotted line in Fig. 5.12 can no longer be normal to the pyramid on the side AB. Vector Y becomes normal to triangle C(δ ) at point 0. Thus the composition of the mixture changes and becomes austenite, (β1 = 0, β2 = 0). The point representing the mixture goes from point C to point D. The education resumes on the side DE following the equation, a

∂ δ Le − (T − Te ) = 0 . ∂t Te

5.5 Education of Shape Memory Alloys

89

Fig. 5.14 The education follows the path (A,C, D, E)

5.5.2.2 Persistence of the Education For the sake of simplicity a schematic pseudopotential has been chosen. It allows a loss of education as it can be seen in (5.31). It is not in agreement with experiments which show that education is often strong enough to persist. In order to agree with experiments, the pseudopotential is sophisticated a little bit by assuming the coefficient a to depend on ∂ δ /dt,

Φ(

a(∂ δ /dt) ∂ δ 2 ∂δ λ , gradT) = (gradT)2 + ( ) , ∂t 2T 2 ∂t

with a(∂ δ /dt) = a for ∂ δ /dt > 0 and a(∂ δ /dt) = a/η where η is small compared to 1, for ∂ δ /dt < 0. It even can be forbidden ∂ δ /dt to become negative by choosing,

Φ(

∂δ λ ∂δ a ∂δ , gradT) = (gradT)2 + ( )2 + I+( ), ∂t 2T 2 ∂t ∂t

where I+ is the indicator function of R+ . It is equivalent to having η = 0. In this last situation, (5.31) describing the evolution becomes ⎞ ⎛ ⎛ ⎞ ⎛ ⎞ −α (T )trε + TL0 (T − T0 ) 0 0 ⎟ ⎝0⎠ ∈ ⎜ ⎠ + ∂ IP (β1 , β2 , δ ), 0 ⎝ α (T )trε + TL0 (T − T0) ⎠ + ⎝ ∂ δ ∂ δ L 0 a ∂ t + ∂ I+ ( ∂ t ) − Te (T − Te ) e

90

5 Shape Memory Alloys

This equation shows it is no longer possible to have the education index δ to evolve when education is achieved. In fact when δ = 1, δ cannot decrease due to ∂ I+ (∂ δ /dt). It cannot increase due to ∂ IP (β1 , β2 , δ ) which makes compulsory for δ to remain equal to 1.

5.5.2.3 Education with Local Microscopic Interaction The equation of motion for the microscopic motions are divH1 − B1 = 0, and divH2 − B2 = 0, in Ω,

(5.33)

with boundary conditions H1 · N = 0, and H2 · N = 0, on ∂ Ω , where N is the outwards normal unit vector to domain Ω occupied by the mixture. Vectors Hi are work flux vectors describing the work fluxes related to the microscopic motions which occur in the martensites and austenite. Equation (5.33) describe the microscopic motions whereas (5.35) below describe the macroscopic motions. For the sake of completeness it is also assumed that there is dissipation with respect to the βi ’s, thus the pseudopotential of dissipation becomes

Φ(

a ∂δ ∂ β1 ∂ β2 ∂ δ λ , , , gradT, T) = (gradT)2 + ( )2 ∂t ∂t ∂t 2T 2 ∂t c 1 ∂ β1 2 c 2 ∂ β2 2 ∂δ ) + ( ) + I+( ), + ( 2 ∂t 2 ∂t ∂t

The equations of motion (5.33) and the constitutive laws give ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ −α (T )trε + TL0 (T − T0 ) c1 ∂∂βt1 − k1 Δ β1 0 ⎟ ⎜ ⎟ ⎝0⎠ ∈ ⎜ ⎝ α (T )trε + TL0 (T − T0 ) ⎠ + ⎝ c2 ∂∂βt2 − k2 Δ β2 ⎠ 0 − LTe (T − Te ) a ∂∂δt + ∂ I+ ( ∂∂δt ) e

+ ∂ IP (β1 , β2 , δ ), in Ω ,

(5.34)

with the boundary conditions

∂ β1 ∂ β2 = 0, and = 0, on ∂ Ω , ∂N ∂N where the ∂ βi /∂ N are the normal derivatives. This equation of motion is to be coupled with the equation for macroscopic motions divσ + f = ργ , in Ω,

(5.35)

5.6 Shape Memory Alloys with the Possibility of Voids

91

where ρ is the density and γ the acceleration, and with the entropy balance if the temperature is assumed not to be known 1 ∂s − λ Δ ln T = ∂t T

  ∂δ ∂ β1 2 ∂ β2 2 a( )2 + c1 ( ) + c2 ( ) , in Ω , ∂t ∂t ∂t

where s is the entropy s=−

∂Ψ L Le d α (T ) = Cc (1 + lnT ) + (β1 + β2) − δ + (β1 − β2 ) trε , ∂T T0 Te dt

giving   d α (T ) ∂ ln T L ∂ β1 ∂ β 2 Le ∂ δ d (β1 − β2 ) + ( + )− + trε ∂t T0 ∂ t ∂t Te ∂ t dt dt   ∂δ ∂ β1 2 ∂ β2 2 1 a( )2 + c1 ( ) + c2 ( ) , in Ω . −λ Δ ln T = (5.36) T ∂t ∂t ∂t

Cc

This entropy equation can be approximated with the small perturbation assumption Cc

∂T ∂ β1 ∂ β2 T0 ∂ δ + L( + ) − Le − λ Δ T = 0, in Ω . ∂t ∂t ∂t Te ∂ t

Let us emphasize again how important is this entropy or energy equation because the thermal actions are equivalent to the mechanical actions. In modelling the behaviour of shape memory alloys the energy balance as to be dealt with the same care and importance as the motions equations. The energy balance is actually important since it has been shown that dissipative phenomena (compared to thermal phenomena) are not as important as it was thought, [64, 65]. A good modelling includes the three coupled equations (5.34), (5.35) with (5.30), and (5.36). One can find examples of thermomechanical evolutions in [210]. Let us note that it is possible to upgrade the theory by assuming local interactions in the education phenomenon, i.e., by having gradδ as a state quantity.

5.6 Shape Memory Alloys with the Possibility of Voids There may be voids in the austenite–austenites mixture, which may appear when the alloy is produced by the aggregation of powders, as it has been recently reported in [195]. Of course, the voids are filled either with gas or air when appearing or when aggregating powders. We do not take into account the gas phase mechanical properties and focus on the mechanical behaviour of the solid mixture, thus we assume the volume fraction of voids is small.

92

5 Shape Memory Alloys

5.6.1 The State Quantities The volume fractions are not independent: they satisfy the following internal constraints 0 ≤ βi ≤ 1,

i = 1, 2, 3,

(5.37)

due to the definition of volume fractions. Since we assume that voids can appear in the martensite–austenite mixture, then the βi ’s must satisfy the other internal constraint

β1 + β2 + β3 ≤ 1,

(5.38)

the quantity vo = 1 − (β1 + β2 + β3 ) being the voids volume fraction. This is the case when the alloy is produced by aggregating powders as shown in [195]. We denote by β the vector of components βi (i = 1, 2, 3) and the set of the state quantities is E = {ε , β , gradβ , T} , while the quantities which describe the evolution and the thermal heterogeneity are

δ E = {ε (U),

dβ dβ , grad , gradT} . dt dt

The gradient, gradβ , accounts for local interactions of the volume fractions at their neighborhood points.

5.6.2 The Mass Balance Assuming the same constant density ρ (see Sect. 4.6 for phase change with different densities, [125]) and the same velocity U for each phase, the mass balance reads

ρ

d(β1 + β2 + β3 ) + ρ (β1 + β2 + β3 )divU = 0 . dt

Within the small perturbation assumption, this equation gives

ρ

∂ (β1 + β2 + β3 ) + ρ (β10 + β20 + β30 )divU = 0, ∂t

5.6 Shape Memory Alloys with the Possibility of Voids

93

where the βi0 ’s are the initial values of the βi . In agreement with the assumption that the voids volume fraction is small, we assume

β10 + β20 + β30 = 1,

(5.39)

∂ (β1 + β2 + β3 ) + divU = 0 . ∂t

(5.40)

and have

Mass balance is a relationship between the quantities of δ E, indeed, its effects will be included in the dissipative properties (see relationship (5.46) below).

5.6.3 The Equations of Motion They result from the principle of virtual power involving the power of the internal forces, −

 Ω

{σ : D(V) + B · γ + H : gradγ } d Ω ,

where V and γ are virtual velocities. The internal forces are the stress σ , the phase change work vector B, and the phase change work flux tensor H. The equations of motion are

ρ

∂U = divσ + f, 0 = divH − B + A, in Ω, ∂t

(5.41)

σ N = g, HN = a, on ∂ Ω ,

(5.42)

where ρ is the density of the alloy which occupies the domain Ω , with boundary ∂ Ω and outward normal vector N. The alloy is loaded by body forces f and by surface tractions g, and submitted to body sources of phase change work A and surfaces sources of phase change work a (for instance, electric, magnetic or radiative actions producing the evolution of the alloys without macroscopic motion). In the following we will suppose, for simplicity, A = a = 0.

5.6.4 The Free Energy As explained above, a shape memory alloy is considered as a mixture of the martensite and austenite phases with volume fractions βi . The volumic free energy

94

5 Shape Memory Alloys

of the mixture we choose is 3

Ψ = Ψ (E) = ∑ βiΨi (E) + h(E),

(5.43)

i=1

where the Ψi ’s are the volume free energies of the i phases and h is a free energy describing interactions between the different phases. We have assumed that internal constraints are physical properties, hence, we decide to choose properly the two functions describing the material, i.e., the free energy Ψ and the pseudopotential of dissipation Φ , in order to take these constraints into account. Since, the pseudopotential describes the kinematic properties (i.e., properties which depend on velocities) and the free energy describes the state properties, obviously the internal constraints (5.37) and (5.38) are to be taken into account with free energy Ψ . For this purpose, we assume the Ψi ’s are defined over the whole linear space spanned by βi and the free energy is defined by

Ψ (E) = β1Ψ1 (E) + β2Ψ2 (E) + β3Ψ3 (E) + h(E) . We choose the very simple interaction free energy k h(E) = IC (β ) + |gradβ |2 , 2 where IC is the indicator function of the convex set C = {(γ1 , γ2 , γ3 ) ∈ R3 ; 0 ≤ γi ≤ 1; γ1 + γ2 + γ3 ≤ 1} .

(5.44)

The terms IC (β ) + (k/2) |gradβ |2 may be seen as a mixture or interaction freeenergy. The only effect of IC (β ) is to guarantee that the proportions β1 , β2 and β3 take admissible physical values, i.e. they satisfy constraints (5.37) and (5.38) (see also (5.44)). The interaction free energy term IC (β ) is equal to zero when the mixture is physically possible (β ∈ C) and to +∞ when the mixture is physically impossible (β ∈ / C). Let us note even if the free energy of the voids phase is 0, the voids phase has physical properties due to the interaction free energy term (k/2) |gradβ |2 which depends on the gradient of β . It is known that this gradient is related to the interfaces properties: gradβ1 , gradβ2 describe properties of the voids–martensites interfaces and gradβ3 describes properties of the voids-austenite interface. In this setting, the voids have a role in the phase change and make it different from a phase change without voids. The model is simple and schematic but it may be upgraded by introducing sophisticated interaction free energy depending on β and on gradβ . For the sake of simplicity, we choose the volume free energies as before

Ψ1 (E) =

1 ε (u) : K : ε (u) − α (T )1 : ε (u) − CT ln T, 2

5.6 Shape Memory Alloys with the Possibility of Voids

95

1 ε (u) : K : ε (u) + α (T )1 : ε (u) − CT ln T, 2 1 L Ψ3 (E) = ε (u) : K : ε (u) − (T − T0 ) − CT ln T, 2 T0

Ψ2 (E) =

where K is the volume elastic tensor and C the volume heat capacities of the phases and the quantity L is the latent heat martensite-austenite volume phase change at temperature T0 as already seen (see Sect. 5.4.5). Concerning the stress α (T )1, we keep the schematic simple expression

α (T ) = (T − Tc )τ 11 , for T ≤ Tc , α (T ) = 0, for T ≥ Tc , with τ 11 ≤ 0 and assume the temperature Tc is greater than T0 . With those assumptions, it results

Ψ (E) =

(β1 + β2 + β3 ) {ε (u) : K : ε (u)} 2 −(β1 − β2 )α (T )1 : ε (u) − β3

L (T − T0 ) T0

k −(β1 + β2 + β3 )CT ln T + |gradβ |2 + IC (β ) . 2

(5.45)

5.6.5 The Pseudopotential of Dissipation The dissipative forces are defined via a pseudopotential of dissipation Φ . As already remarked, the mass balance (5.40) is a relationship between velocities of δ E. Thus we take it into account in order to define the pseudopotential and introduce the indicator function I0 of the origin of R as follows I0 (

∂ (β1 + β2 + β3 ) + divU) . ∂t

From experiments, it is known that the behaviour of shape memory alloys depends on time, i.e., the behaviour is dissipative. We define a pseudopotential of dissipation with χ = T 

Φ

 c ∂ β 2 υ ∂β ∂β ∂β 2 , grad , gradT, T = | + |grad ∂t ∂t 2 ∂t 2 ∂t +

λ ∂ (β1 + β2 + β3 ) |gradT|2 + I0 ( + divU), 2T ∂t

(5.46)

96

5 Shape Memory Alloys

where λ ≥ 0 represents the thermal conductivity and c ≥ 0, υ ≥ 0 stand for phase change viscosities. We have assumed dissipation with respect to grad(∂ β /∂ t) to show how it intervene.

5.6.6 The Constitutive Laws The internal forces are split between non-dissipative forces σ nd , Bnd and Hnd depending on (E, x,t) and dissipative forces by

σ d , Bd , Hd , −Q ,

depending on

δE = {

∂ε ∂β ∂β , , grad , gradT}, ∂t ∂t ∂t

(5.47)

and on (E, x,t)

σ = σ nd + σ d , B = Bnd + Bd , H = Hnd + Hd . The non dissipative forces are defined with the free energy

σ nd (E) =

∂Ψ (E) = (β1 + β2 + β3 )K : ε (u) − (β1 − β2 )α (T )1, (5.48) ∂ ε (u)

Bnd (E, x,t) = ⎛

∂Ψ (E) ∂β

⎞ ε (u) : K : ε (u) − 2α (T )1 : ε (u) 1⎜ ⎟ = ⎝ ε (u) : K : ε (u) + 2α (T )1 : ε (u) ⎠ + Bndr (E, x,t), 2 ε (u) : K : ε (u) − 2 TL0 (T − T0 )

(5.49)

Bndr (E, x,t) ∈ ∂ IC (β ),

(5.50)

Hnd =

∂Ψ (E) = kgradβ . ∂ (gradβ )

The dissipative forces are defined with the pseudopotential of dissipation

σ d , Bd , Hd , −Qd ∈ ∂ Φ (δ E, χ ),

(5.51)

(5.52)

where the subdifferential of Φ is with respect to δ E. Relationship (5.52) gives

σ d = −p1,

(5.53)

5.6 Shape Memory Alloys with the Possibility of Voids

97

⎛ ⎞ 1 ∂β d ⎝ , B = −p 1 ⎠ + c ∂t 1 Hd = υ grad −Qd =

(5.54)

∂β , ∂t

(5.55)

λ gradT, T

(5.56)

where p is the pressure in the mixture and defined by − p ∈ ∂ I0 (

∂ (β1 + β2 + β3 ) + divU) . ∂t

(5.57)

The state laws (5.48)–(5.51), besides implying that the internal constraints are satisfied, give also the value of the reactions, during the evolution, to these internal constraints. Relationships (5.48)–(5.51) and (5.53)–(5.57) give the constitutive laws

σ = (β1 + β2 + β3 )K : ε (u) − ((β1 − β2)α (T ) + p)1, ⎛

ε (u) : K : ε (u) − 2α (T )1 : ε (u) − p

(5.58) ⎞

⎜ ⎟ ⎟ 1⎜ ε (u) : K : ε (u) + 2 α (T )1 : ε (u) − p ⎜ ⎟ B(E, δ E, x,t) = ⎜ ⎟ 2⎝ ⎠ L ε (u) : K : ε (u) − 2 (T − T0 ) − p T0

∂β , ∂t

(5.59)

Bndr (E, x,t) ∈ ∂ IC (β ),

(5.60)

+Bndr (E, x,t) + c

−p ∈ ∂ I0 (

∂ (β1 + β2 + β3 ) + divU), ∂t

(5.61)

∂β , ∂t

(5.62)

H = kgradβ + υ grad

λ −Q(E, δ E) = −Qd (E, δ E) = − gradT . T

(5.63)

It can be proved that our choice is such that the internal constraints and the second law of thermodynamics are satisfied, (see [112, 119] and the next Sect. 5.6.7).

98

5 Shape Memory Alloys

5.6.7 The Entropy Balance By denoting s=−

∂Ψ L = (β1 + β2 + β3 )C(1 + lnT ) + β3 , ∂T T0

(5.64)

the entropy balance is 

 ∂ε ∂β d ∂β d +B · + H : grad − Q · gradT σ : ∂t ∂t ∂t   1 ∂β 2 ∂β 2 λ 2 = R+ c| | + υ |grad | + |gradT| , in Ω , (5.65) T ∂t ∂t T

1 ∂s + div Q = R + ∂t T

d

−QN = π , on ∂ Ω ,

(5.66)

because  p

 ∂ (β1 + β2 + β3 ) + divU = 0, ∂t

due to (5.57), the pressure is workless. Vector T Q is the heat flux vector, RT is the exterior volume rate of heat that is supplied to the alloy, T π is the rate of heat that is supplied by contact action, ∂ ε /∂ t = ε (U) is the strain rate. The constitutive laws, within the small perturbation assumption and (5.39), become

σ = K : ε (u) − ((β1 − β2)α (T ) + p)1,

(5.67)

∂ (β1 + β2 + β3 ) + divU), −p ∈ ∂ I0 ( ∂t ⎛ ⎞ −α (T )1 : ε (u) − p ⎜ α (T )1 : ε (u) − p ⎟ ⎟ + Bndr + c ∂ β , B=⎜ ⎝ ⎠ L ∂t − (T − T0 ) − p T0 Bndr ∈ ∂ IC (β ),

∂β , ∂t λ Q(E, δ E) = Qd (E, δ E) = − gradT . T H = kgradβ + υ grad

(5.68)

The entropy becomes s=−

∂Ψ L = C(1 + ln T ) + β3 , ∂T T0

(5.69)

5.7 Conclusion

99

5.6.8 The Set of Partial Differential Equations We assume also quasi-static evolution and, using again the small perturbation assumption (i.e. neglecting the higher order contributions in the velocities in (5.65) which are smaller than the other quantities), we get the following set of partial differential equations coupling the equations of motion (5.41), the entropy balance (5.65) with entropy (5.69) and constitutive laws (5.67)–(5.68) div (K : ε (u) − ((β1 − β2 )α (T ) + p)1) + f = 0,

(5.70)

∂ (β1 + β2 + β3 ) + divU), ∂t ⎛ ⎞ −α (T )1 : ε (u) − p ∂β ∂β ⎜ ⎟ − υΔ − kΔ β + ⎝ α (T )1 : ε (u) − p ⎠ + Bndr = 0, c ∂t ∂t − TL0 (T − T0 ) − p

−p ∈ ∂ I0 (

Bndr ∈ ∂ IC (β ), C

∂ ln T L ∂ β3 + − λ Δ ln T = R . ∂t T0 ∂ t

This set is completed by suitable initial conditions and the following boundary conditions:

σ N = g, on Γ1 ,

(5.71)

u = U = 0, on Γ0 ,

υ

∂β ∂ ∂β ( )+k = 0, on ∂ Ω , ∂N ∂t ∂N ∂ (ln T ) = π , on ∂ Ω , λ ∂N

where g is the exterior contact force applied to Γ1 , where (Γ0 , Γ1 ) is a partition of ∂ Ω . It may be proved that this set of partial differential equations has solutions in a convenient mathematical setting, insuring the model is coherent from both the mathematical and the mechanical point of view allowing numerical approximations, [126].

5.7 Conclusion The evolution of a structure made of shape memory alloys, i.e., the computation of E(x,t) = (ε (x,t), β1 (x,t), β2 (x,t), β3 (x,t), gradβ1 (x, t), gradβ2 (x, t), gradβ3 (x, t), T(x, t)),

100

5 Shape Memory Alloys

together with the small displacement u(x,t) depending on the point x of the domain occupied by the structure and on time t, can be performed by solving numerically the set of partial differential equations, (5.27)–(5.29), resulting from the equations of motion, the entropy balance and the constitutive laws (5.24)–(5.26), completed by convenient initial and boundary conditions (or equations of motion (5.41), (5.42), entropy balance (5.65), (5.66) and constitutive laws (5.58)–(5.63)), [36, 37, 44, 63, 69, 74, 112, 179, 210]. The models we have described are able to account for the different features of the shape memory alloy macroscopic, mechanical and thermal properties. We have used schematic free energies and schematic pseudopotentials of dissipation. There are many possibilities to upgrade the basic choices we have made to take into account the practical properties of shape memory alloys. Let us, for instance, mention that the pseudopotential of dissipation can be modified to describe more precisely the hysteretic properties. There is no difficulty in having more than two martensites, for instance, to take care of 24 possible martensites! In the same way, it is possible to introduce different forms one martensite may exhibit, as in [156]. Note that the physical quantities for characterizing an educated shape memory alloys are K, C, L, T0 , Tc , τ , the two martensite volume fractions, coordinates of point Be defining triangle Ce , for the free energy and c, k, λ for the pseudopotential of dissipation. Those are not so many to have a complete multidimensional model which can be used for engineering purposes. Let us note the very important role of internal constraints and of the reaction Bndr to those internal constraints which are responsible for many properties. Let us also note that when there are voids, the pressure is the reaction to the kinematic constraint resulting from the mass balance. From this point of view, pressure p is involved in the equations in a logic and clear way.

Chapter 6

Damage

It is known that damage results from microscopic motions in the structures, as it is caused by microfractures and microcavities resulting in the decreasing of the material stiffness. There is wide literature on this topic and it is difficult to be exhaustive. Let us mention [145, 151, 152] modelling damage within the framework of continuum mechanics. The damage quantities are internal quantities which are involved in the free energy of materials [104, 129, 138, 151, 152]. The many possible expressions for the free energy yield numerous, versatile constitutive laws [4,7,9,26,80,138,151,196]. The damage quantities may also appear in the equations of motions [50,75,104,107,112,120,137,148,165,175]. They may also approximate fractures [6, 52, 53, 77, 81, 100–102, 176]. Predictive theories of damage of living materials are also developed, [163]. At the macroscopic level, damage is represented by the state quantity β with values between 0 and 1. When β = 0, the material is completely damaged, when β = 1, the material is sound or undamaged. When 0 < β < 1, the material is partly damaged. State quantity β may be thought as the volume fraction of microfractures and microcavities. In many practical situations, the evolution is irreversible. But there are cases where it is not. For instance, some polymers may recover there strength if left at rest after a damaging loading. This is also the case for some bituminous materials which damage during hot days and mend by themselves during cool nights. The state quantities we choose are E = (T, ε , β , gradβ ), where ε = ε (u) is the small deformation and u the small displacement. The gradient of the volume fraction is introduced to take into account local interactions or the influence of a material point on its neighbourhood. What are the velocities we need to describe the motion? We have the macroscopic velocities U. But we have said that damage results from microscopic motions, thus we have to account for these motions. Let us remark that when there are microscopic motions, the damage quantity β evolves and when there are not microscopic motions, β remains constant. M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 6, © Springer-Verlag Berlin Heidelberg 2012

101

102

6 Damage

Thus we choose to account at the macroscopic level for the microscopic motions by velocity dβ . dt The actual or virtual velocities have two components (U(x,t), (dβ /dt)(x,t)) and (V(x), γ (x)) where V, γ are macroscopic and microscopic virtual velocities. By stretch of language we may say that dβ /dt is the microscopic velocity whereas we should say that it is the damage velocity accounting for the microscopic velocity. The theory has been established in Chaps. 2 and 3 with a different physical meaning for β .

6.1 The Equations of Motion The equations of motion established in Chaps. 2 and 3 with a different physical meaning for β are

ρ ρ0

dU − divσ = f, in Ω , σ N = g, on ∂ Ω , dt

d2 β − divH = A − B, in Ω , H · N = a, on ∂ Ω . dt 2

(6.1)

The second equation of motion (6.1) is new. It accounts for the microscopic motions. Both the partial differential equation and the boundary condition have a precise physical meaning: they describe how work is provided to the structure without macroscopic motion. In the sequel, we consider only quasi-static evolutions. Results involving the accelerations are reported in [49, 123].

6.2 Free Energy and Pseudopotential of Dissipation The free energy and pseudopotential of dissipation with χ = T , we choose, are assuming small perturbations

Ψ (ε , β , gradβ , T) = −CT ln T +

β {λe (trε )2 + 2μe ε : ε } + w(1 − β ) 2

k +I(β ) + (gradβ )2 , 2   c ∂β 2 ∂β λ ∂β , T) = (gradT)2 + ), Φ (gradT, + I− ( ∂t 2T 2 ∂t dt

6.3 The Equations

103

where λe and μe are the elasticity Lam´e parameters and, c is the viscosity of damage, i.e., the viscosity of the microscopic motions, k is the microscopic interaction coefficient and w is the cohesion energy. The constitutive are

σ = β {λe (trε )1 + 2 μeε } , ∂β ∂β 1 {λe (trε )2 + 2 μe ε : ε } − w + c + ∂ I(β ) + ∂ I−( ), H = kgradβ , 2 ∂t ∂t λ Q = − gradT, T s = C(1 + ln T ). B∈

Let us note that the actual Lam´e parameters are proportional to β : β λe and β μe . Thus the Young modulus is proportional to β . The linear function w(1 − β ) in free energy, is responsible for a threshold in the constitutive law for the damage. Depending of the value of the elastic energy (1/2){λe (trε )2 + 2μe ε : ε } with respect to w, damage occurs or does not occur, as it results from the equation of microscopic motion. Information on the measurements of the physical parameters may be found in [175].

6.3 The Equations The equation of the predictive theory are

∂ ln T − λ Δ ln T = R, in Ω , ∂t ∂β ∂β − kΔ β + ∂ I(β ) + ∂ I−( ) c ∂t ∂t 1  w − {λe (trε )2 + 2μe ε : ε } + A, in Ω , 2 div (β (λe (trε (u)1 + 2μeε (u))) + f = 0, in Ω,

C

λ

(6.2) (6.3)

∂ ln T ∂β = π, k = a, σ N = g, on ∂ Ω , ∂N ∂N

T (x, 0) = T 0 (x), β (x, 0) = β 0 (x), in Ω . The damage sources A and a take into account chemical, electrical, radiative actions, [21, 24], which can break the links insuring the cohesion of the material. Note that in this setting the thermal problem is decoupled from the mechanical problem. The thermal effects of damage due to the dissipation are investigated in [142]. Mechanical results may be found in [112, 113, 120–122, 168, 175, 180] and

104

6 Damage

mathematical results in dimension one may be found in [117,118], and in dimension two and three in [46, 116].

6.4 The Macroscopic Motions Become Microscopic In this predictive theory, there are relationships between macroscopic and microscopic motions during the damaging process. Indeed, as the model is based on the separation between the description of macroscopic and microscopic motions, it is natural to investigate what occurs when macroscopic motions vanish progressively and become microscopic. In order to answer this question let apply exterior actions to a structure in such a way that the amplitude of the resulting macroscopic motions become smaller and smaller un → 0, but the strain rates are highly oscillating and to not tend to 0 1 {λe (trε (un ))2 + 2μe ε (un ) : ε (un )} → d ≥ 0, 2 in some weak sense. The macroscopic motions vanish and it is no longer possible to consider that they are macroscopic: one has to consider them as microscopic. One may wonder if their damaging effects also vanish. If they do not, how are they taken into account by the theory? Is there a transfer from (6.3) which describes the macroscopic motions towards (6.2) which describes the microscopic motions? The theory shows that when macroscopic motions become microscopic, their damaging effects remain as a source of damage d, which is clearly related to microscopic motions, [42, 113, 168]. The limit equations for β = lim β n and u = lim un = 0, are

∂ ln T − λ Δ ln T = R, in Ω , ∂t ∂β ∂β c − kΔ β + ∂ I(β ) + ∂ I−( )  w − d, in Ω , ∂t ∂t u = 0, C

T (x, 0) = T 0 (x), β (x, 0) = β 0 (x), in Ω . The work which is provided to the structure by the macroscopic forces becomes the work provided by the damage source d. In the limit, the rapid oscillations with vanishing amplitude are no longer seen because u = 0, but their effect is still present via the external damage source d. We conclude that the damaging effect of a vanishing macroscopic motion does not

6.4 The Macroscopic Motions Become Microscopic

105

Fig. 6.1 The wedge spitting test. The progressive concrete damage. The color scales are slightly different in the pictures. In the red domains, damage is at its maximum which is different in each picture. The green domains are not damaged

whl

u = 0.1 mm

M5F side A 2

4

1

3

Endommagement .218415 .609207 .296573 .687366 .374732 .765524 .45289 .843683 .531049 .921841 .609207 1.

u

Fig. 6.2 The wedge splitting test experiment. A displacement is applied as shown on the figure. A splitted sample and the value of damage in the sample when displacement is 0.1 mm

106

6 Damage

u = 0.3 mm

Endommagement .0869802 .177979 .268977 .359975 .450974 .541972

.541972 .632971 .723969 .814967 .905966 .996964

Fig. 6.3 The applied displacement is 0.3 mm. The damage is more important than for the 0.1 mm. applied displacement. The displacements in the damage zone are important and account for the fracture

Fig. 6.4 The wedge splitting experiment. Force versus displacement is plotted. The points are experimental results. The continuous line are numerical results for different meshes, [175]

vanish. It is taken into account by the equation of microscopic motion via a damage source whose intensity is the work provided to establish the vanishing motion. One can also say that the damage source includes the chemical, radiative, electrical,... actions, as already mentioned, but also the mechanical actions which are not accounted for by the macroscopic displacement u, for instance rapid microscopic

6.5 An Example: The Wedge Splitting Test

107

oscillations. In this case the external action A includes the work d provided to the structure by the rapid oscillations.

6.5 An Example: The Wedge Splitting Test The wedge splitting experiment is shown in Figs. 6.1 and 6.2, together with the picture of samples after splitting into two pieces, [134]. Some numerical results due to Francesco Freddi are shown in Figs. 6.1, 6.2 and 6.3. Experimental and numerical results due Boumediene Nedjar are shown in Fig. 6.4, [175]. The softening of the curve force versus-displacement result from damage, i.e., from the decrease of β .

•

Chapter 7

Contact with Adhesion

Let us consider two solids Ω1 and Ω 2 glued on one another on their contact surface Γ . In order to take into account the adhesive properties of the glue which result from fibers connecting the contact surfaces and breaking progressively, Fig. 7.1, [57, 79], we choose as state quantity the surface fraction of active glue fibers, β (x,t), [31,58, 67,76,82,104,106–108,168,180,187,204,206]. When β (x,t) = 0 all the glue fibers are broken and Signorini contact properties are valid: the interactions of the two solids result only from the impenetrability condition, [11, 98, 161, 182, 199]. When β (x,t) = 1, the glue is active: it prevents the separation when tension is applied. The interactions between the two solids result both from the impenetrability condition and the adhesion properties. When β (x,t) is between 0 and 1, part of the glue fibers are active. The glue is partly damaged: the interaction between the two solids is in between the two extreme interactions. The evolution of the surface fraction of active fibers is produced by microscopic motions which break or mend the glue fibers. We think that the power of these microscopic motions are to be taken into account in a predictive theory. The previous damage example and theory of Sect. 3.1 have shown how to build a macroscopic theory: we choose dβ /dt on the contact surface to account at the macroscopic level for the microscopic motions. The state quantities of the system made of the two solids are besides β , the two small deformations ε1 , ε2 and the gap u2 − u1 on the contact surface Γ , where the ui are the small displacements of the two solids. For the sake of simplicity, we neglect the thermal phenomena and do not choose the temperature as a state quantity. As well as for the volume damage, there is an abundant bibliography on contact with adhesion, involving peeling [25,166,167,184,186]. Let us also mention results on damage of dimension two solids, [205].

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 7, © Springer-Verlag Berlin Heidelberg 2012

109

110

7 Contact with Adhesion

Fig. 7.1 Peeling experiment: long fibers connect two solids: a flat one and a thin one which is slightly grey. The fibers are shiny. Experiment of M. Horgnies, [79]

7.1 The Equations of Motion They result from the principle of virtual power which involves the virtual power of the internal forces Pint (V1 , V2 , γ ) = − −

 Ω1



Γ

σ1 : D(V1 )d Ω 1 −

 Ω2

σ2 : D(V2 )d Ω2



(Bγ + H · grads γ )dΓ −

Γ

R · (V2 − V1 )dΓ,

where V1 , V2 are macroscopic virtual velocities of the two solids, γ is a microscopic virtual velocity and grads is the surface gradient. The force R is the action of solid 1 on solid 2 on contact surface. This vector describes for instance the friction forces. The virtual power of the exterior forces is Pext (V1 , V2 , γ ) =

 Ω1

+ +

f1 · V1 d Ω1 +

 

∂ Ω 1 \Γ

Γ

 Ω2

f2 · V2 d Ω2

g1 · V1 d Γ +

Aγ d Γ +

 ∂Γ

 ∂ Ω2 \Γ

g2 · V2 dΓ

aγ dΓs ,

where the fi and gi are the body and surface forces applied to the solids. The A and a are respectively the surface and line exterior sources of microscopic work which

7.2 The Constitutive Laws

111

can damage the glue (the line ∂Γ is the boundary of the surface Γ ). It is known that chemical action and optical actions damage glue (light damages some type of glue). It is assumed these works are known. The virtual work of acceleration forces is Pacc (V1 , V2 , γ ) =

 Ω1

ρ1

dU1 · V1 d Ω1 + dt

 Ω2

ρ2

dU2 · V2 d Ω 2 + dt

 Γ

ρ0

d2 β γ dΓ , dt 2

where ρ is the density of the solid and ρ0 is proportional to the density of the glue. The principle of virtual power ∀ V1 , V2 , γ , Pacc (V1 , V2 , γ ) = Pint (V1 , V2 , γ ) + Pext (V1 , V2 , γ ), gives easily the equations of motion dU1 − divσ1 = f1 , in Ω1 , σ1 N1 = g1 , on ∂ Ω1 , dt dU2 − divσ2 = f2 , in Ω2 , σ2 N2 = g2 , on ∂ Ω2 , ρ2 dt σ1 N1 − R = 0, σ2 N2 + R = 0,

ρ1

ρ0

d2 β + B − divs H = A, on Γ , H · ns = a, on ∂Γ , dt 2

(7.1)

where the vector Ni is the outside normal vector to Ωi , ns is the surface outside vector to Γ and divs is the surface divergence.

7.2 The Constitutive Laws They result from volume and surface free energies and pseudopotential of dissipation. We choose

Ψi (εi ) = Φi (

1 {λi (trεi )2 + 2μiεi : εi }, 2

∂ εi ) = 0, ∂t

for the volume functions. The small deformations εi due to small displacement ui are 1 εi = ε (ui ) = (gradui + (gradui )T ), 2 where (gradui )T is the transposed matrix of gradui . Thus each solid is elastic. We choose

112

7 Contact with Adhesion

k ks β (u2 − u1 )2 + ws (1 − β ) + (grads β )2 2 2 +I(β ) + I− ((u2 − u1 ) · N2 ) ,   ∂β ∂β cs ∂ β 2 )= ), Φ( + I− ( ∂t 2 ∂t ∂t

Ψ (u2 − u1 , β , grads β ) =

for the contact surface functions. Function I− is the indicator function of R− , taking into account the impenetrability of the two solids (u2 − u1 ) · N2 ≤ 0. This contact surface has an elastic reaction whose rigidity is proportional to β . The evolution of the glue is dissipative with coefficient of viscosity cs and, irreversible: once the glue fibers are broken they do not mend by themselves. This irreversible behaviour characterizes solid glue. In case we consider fresh and liquid glue, the glue fibers once broken are rebuilt if the solids get close to one another. In this situation, the evolution is reversible and the indicator function I− (∂ β /dt) is removed from the surface pseudopotential. The cohesion of the glue, called the Dupr´e’s energy [166], is ws . The non dissipative local interactions of the glue fibers is characterized by ks . The constitutive laws are

σi = λi (trεi )1 + 2μiεi , in the solids. On the contact surface the constitutive laws are R ∈ kβ (u2 − u1 ) + ∂ I− ((u2 − u1 ) · N2 ) N2 ,

∂β ∂β k B ∈ (u2 − u1 )2 − ws + cs + ∂ I(β ) + ∂ I−( ), 2 ∂t ∂t H = ks grads β .

7.3 The Equations We assume that no work involving microscopic motions is provided to the system, i.e., there are no chemical, radiative, optical or electrical actions: A = 0 and a = 0. We assume solid 2 is fixed to a rigid support on part Γ2 . The equations of the predictive theory for a quasi-static evolution are

7.4 Examples

k ∂β ∂β − ks Δ s β + ∂ I(β ) + ∂ I−( )  ws − (u2 − u1 )2 , on Γ , ∂t ∂t 2 div {λi trε (ui )1 + 2 μiεi (ui )} + fi = 0, in Ωi ,

cs

113

(7.2) (7.3)

σ1 N1 = g1 , on ∂ Ω 1 \Γ , σ2 N2 = g2 , on ∂ Ω 2 \(Γ ∪ Γ2 ), u2 = 0, on Γ2 , σ1 N1 − kβ (u2 − u1 ) − ∂ I− ((u2 − u1 ) · N2 ) N2  0, on Γ , σ2 N2 + kβ (u2 − u1 ) + ∂ I− ((u2 − u1 ) · N2 ) N2  0, on Γ , k

∂β = 0, on ∂Γ , ∂ ns

β (x, 0) = β 0 (x), on Γ , where β 0 is the initial surface fraction of unbroken glue fibers. For instance, β 0 = 1 when the glue is sound. The elements of the subdifferential set ∂ I− ((u2 − u1 ) · N2 ) are the impenetrability reactions which are active only when normal gap (u2 − u1 ) · N2 is null.

7.4 Examples They are given in the next chapter where volume damage is coupled to surface damage.

•

Chapter 8

Damage of Solids Glued on One Another: Coupling of Volume and Surface Damages

Let us consider two pieces of concrete glued on one another. Both the concrete and the glue responsible for the adhesion, can damage due to external actions. The volume and surface damages are coupled. The interesting part of the predictive theory is the set of equations on the contact surface, [99, 100]. Mathematical results are reported in [43].

8.1 State Quantities and Quantities Describing the Evolution We neglect the thermal effects and do not take the temperature into account. The state quantities and quantities describing the evolution are: in Ω1 and in Ω2   ∂ ε1 ∂ β1 ∂ β1 , , , grad E1 = {ε1 , β1 , grad β1 }, δ E1 = ∂t ∂t ∂t   ∂ ε2 ∂ β2 ∂ β2 E2 = {ε2 , β2 , grad β2 } , δ E2 = , , , grad ∂t ∂t ∂t where the ε ’s are the small deformations and the β ’s the volume damages; in ∂ Ω1 ∩ ∂ Ω2 Es = {u2 − u1 , βs , grads βs , β1 , β2 },   ∂ βs ∂ β s ∂ β1 ∂ β 2 , , grads , , δ Es = U2 − U1 , ∂t ∂t ∂t ∂t with U1 =

∂ u1 ∂ u2 , U2 = , ∂t ∂t

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 8, © Springer-Verlag Berlin Heidelberg 2012

115

116

8 Damage of Solids Glued on One Another

where the u’s are the small displacements with velocities U = ∂ u/dt, βs is the surface or glue damage. On the contact surface, we use the gap u2 − u1 and its velocity as it is usual in contact mechanics. But we will have to deal with damage resulting from the elongation or the stretching of the contact surface. In an elongation where ∀x, u2 (x) − u1 (x) = 0, the gap is 0 because the two displacements are equal. Thus we choose to introduce a non local deformation quantity g(y, x) = 2(y − x) · (u2 (y) − u1 (x)), which does not vanish in an elongation. Es,1,2 = {g(y, x) = 2(y − x) · (u2 (y) − u1 (x)), βs (x), βs (y)},   ∂ βs ∂ βs (x), (y) , δ Es,1,2 = D1,2 (U1 , U2 ), ∂t ∂t with D1,2 (U1 , U2 )(x, y) = 2(y − x) · (U2 (y) − U1 (x)). The last quantity is the velocity of the non local deformation g(y, x). Let us note that this velocity is 0 in any rigid body velocity, V + ϖ × x. This property is obvious because D1,2 is the time derivative of the square of the distance of two material points d D1,2 (U1 , U2 )(x, y) = (y(t) − x(t))2 . dt

8.2 Equations of Motion They result from the principle of virtual power which involves the powers of the interior forces, of the exterior forces and of the acceleration forces, [112].

8.2.1 Virtual Power of the Interior Forces Both volume damage and surface damage result from microscopic motions whose power is taken into account in the power of the interior forces. We choose the velocities ∂ β /dt to account for the microscopic velocities at the macroscopic level. In order to take into account local interactions, we introduce the gradients grad(∂ β /dt). Assuming V = (V1 , V2 ) and γ = (γ1 , γ2 , γs ) to be macroscopic and microscopic virtual velocities, the virtual power of the interior forces, which is linear function of the virtual velocities, is chosen as

8.2 Equations of Motion

Pint (V, γ ) = − − − − + +

 Ω1



Ω2

σ1 : D(V1 )d Ω − σ2 : D(V2 )d Ω −



∂ Ω1 ∩∂ Ω 2



∂ Ω1 ∩∂ Ω 2



117

Ω1



Ω2

{B1 γ1 + H1 · grad γ1 } d Ω {B2 γ2 + H2 · grad γ2 } d Ω

R · (V2 − V1 )dΓ {Bs γs + Hs · grads γs + B1,s (γ1 − γs ) + B2,s(γ2 − γs )} dΓ 

∂ Ω1 ∩∂ Ω 2 ∂ Ω 1 ∩∂ Ω2







∂ Ω1 ∩∂ Ω 2 ∂ Ω 1 ∩∂ Ω2

M(x, y)D1,2 (V1 , V2 )(x, y)dxdy {Bs,1 (x, y)γs (x) + Bs,2(x, y)γs (y)} dxdy.

The different quantities which contribute to the power of the interior forces are products of kinematic quantities by interior forces. The kinematic quantities are the quantities that intervene in the motion we intend to describe. Their choice is of paramount importance since they determine the whole predictive theories. They are chosen by experimenting phenomena which are volume and surface deformations together with volume and surface damage, i.e., micro-voiding and micro-cracking. Thus there are quantities with surface and volume densities depending on the quantities we have chosen to describe the evolutions or the deformations of the system. Some of them are classical others are new. Most of them are local but a few are non local because there is a kinematic quantity which is non local. Let us comment on the different power densities: • The usual strain rate D introduces the stress σ . • The damage velocity, ∂ β /dt is a scalar, thus the associated interior force is also a scalar, B. It is a work, the internal damage work which is responsible for the evolution of the damage in the volume and in the surface. • The gradient of the damage velocity, grad(∂ β /dt) is a vector, thus the interior force is a vector, H. It is a work flux vector which is responsible for the interaction of the damage at a point on the damage of its neighbourhood. Its physical meaning is to be given by the boundary condition of the equation of motion as the physical meaning of the stress is given by the boundary condition of the equation of motion. • The gap velocity U2 − U1 on the contact surface introduces the classical macroscopic interaction force R. • In the same way the difference between the damage velocities ∂ βi /dt − ∂ βs /dt introduces a damage work flux on the surface, Bi,s which describes the interaction of volume damage and surface damage. • The elongation velocity, D1,2 (U1 , U2 )(x, y) which is non local, introduces a non local scalar M(x, y) interior force. It describes the effects of the elongation. It results in the equations of motion as a classical force. The interaction

118

8 Damage of Solids Glued on One Another

macroscopic mechanical force has a non local part and a classical local part, the force R (see formula (8.2) below). Since we are going to assume the interior force M(x, y) depending on the surface damage βs , it is wise to add an extra non local power depending on damage velocity ∂ βs /dt. It describes the effect of damage at point x on damage at point y. The interior forces Bs,i (x, y) have the same effect than M: they introduce a non local internal source of damage work. The microscopic mechanical force has a non local part and three local parts, Bs due to the glue and the two Bi,s due to the interactions with the volumes (see formula (8.1) below). Let us note that even if the interior forces are numerous and some of them are unusual, all of them are simple and precisely committed to take into account a particular aspect of the coupling of volume and surface, and of microscopic and macroscopic evolution of the system.

8.2.2 Virtual Power of the Exterior Forces We assume no exterior microscopic, either surface or volume, source of damage such as radiative, electrical or chemical damaging actions. Thus we have Pext (V, γ ) =

 Ω1

+

f 1 · V1 d Ω +

 Ω2

 ∂ Ω1 \(∂ Ω1 ∩∂ Ω 2 )

f 2 · V2 d Ω +

g1 · V1 d Γ



∂ Ω2 \(∂ Ω1 ∩∂ Ω2 )

g 2 · V2 d Γ ,

where the f and g are the body and surface exterior forces.

8.2.3 Virtual Power of the Acceleration Forces For the sake of simplicity, we assume a quasi-static problem. Thus Pacc (V, γ ) = 0.

8.2.4 The Principle of Virtual Power It is ∀V = (V1 , V2 ), γ = (γ1 , γ2 , γs ), Pacc (V, γ ) = Pint (V, γ ) + Pext (V, γ ).

8.3 The Constitutive Laws

119

8.2.5 The Equations of Motion They result from the principle. By choosing convenient virtual velocities, we get easily for domain Ω1 div σ1 + f1 = 0, −B1 + divH1 = 0, in Ω1 ,

σ1 N1 = g1 , H1 · N1 = 0, in ∂ Ω1 \(∂ Ω 1 ∩ ∂ Ω 2 ). They are the volume equations of motion accounting for macroscopic and microscopic effects. The equations of motion for domain Ω 2 are the same. For the sake of simplicity, we consider the domains Ω1 and Ω2 have the same mechanical properties and give only the mechanical relationships for domain Ω1 . The related boundary conditions and the equation of motion on ∂ Ω1 ∩ ∂ Ω2 involve non local forces.

σ1 N1 (x) = R(x) +

 ∂ Ω1 ∩∂ Ω 2

σ2 N2 (y) = −R(y) +

2(x − y)M(x, y)dy, x ∈ ∂ Ω1 ∩ ∂ Ω2 ,



∂ Ω1 ∩∂ Ω 2

2(y − x)M(x, y)dx, y ∈ ∂ Ω 1 ∩ ∂ Ω2 ,

H1 · N1 = −B1,s , H2 · N2 = −B2,s , −Bs (x) + divs Hs (x) + B1,s(x) + B2,s(x) −



∂ Ω1 ∩∂ Ω 2

Bs,1 (x, y) + Bs,2 (y, x)dy = 0, x ∈ ∂ Ω 1 ∩ ∂ Ω2 , (8.1)

Hs · ns = 0, on ∂ (∂ Ω 1 ∩ ∂ Ω 2 ), where ns is the normal vector to the boundary ∂ (∂ Ω1 ∩ ∂ Ω2 ) of ∂ Ω 1 ∩ ∂ Ω2 . Constitutive laws are needed for the numerous interior forces. As usual, we choose to define them with free energies depending on the state quantities E and pseudopotential of dissipation depending on the velocities δ E. Remark 8.1. Function M(x, y) is not symmetric. Because there are at a distance interactions, there is no clear difference between the contact surface considered as a structure and contact surface considered as a material. In case a subdomain of ∂ Ω1 ∩ ∂ Ω2 is considered, the non local actions result in interior non local actions and exterior non local actions. This point of view is developed in [112].

8.3 The Constitutive Laws Because the thermal phenomenon are not taken into account, the second law of thermodynamics is

120

8 Damage of Solids Glued on One Another

dΨ1 ∂ β1 ∂ β1 (ε1 , β1 , grad β1 ) ≤ σ1 : D(U1 ) + B1 + H1 · grad , in Ω1 , ∂t ∂t ∂t ∂ βs dΨs (u2 − u1 , βs , grads βs , β1 − βs , β2 − βs) ≤ R · (U2 − U1 ) + Bs ∂t ∂t ∂ βs ∂ β1 ∂ βs ∂ β2 ∂ βs +Hs · grads + B1,s ( − ) + B2,s( − ), on ∂ Ω 1 ∩ ∂ Ω 2 , ∂t ∂t ∂t ∂t ∂t dΨs,1,2 (g(y, x), βs (x), βs (y)) ≤ −M(x, y)D1,2 (U1 , U2 )(x, y) ∂t ∂ βs ∂ βs (x) − Bs,2(x, y) (y), in (∂ Ω1 ∩ ∂ Ω2 ) × (∂ Ω1 ∩ ∂ Ω2 ). −Bs,1 (x, y) ∂t ∂t These relationships are used to define the constitutive laws with pseudopotential of dissipation. The + or − signs appearing in the constitutive laws result from the + or − signs which are in the right hand sides of the inequalities. The right hand sides are the opposite of the densities of the actual powers of the interior forces. The free energy and pseudopotential of dissipation are the same for each domain. For domain Ω1 , they are

Ψ1 (E1 ) = Ψ1 (ε1 , β1 , grad β1 ) = w1 (1 − β1) +

k1 (grad β1 )2 + I(β1 ) 2

 β1  λ1 (trε1 )2 + 2μ1 ε1 : ε1 , 2       ∂ β1 ∂ β1 c 1 ∂ β1 2 = , Φ1 (δ E1 ) = Φ1 + I− ∂t 2 ∂t ∂t +

where w1 is the damage threshold, k1 the damage interaction parameter which quantifies the influence of volume damage on its neighbourhood, λ1 and μ1 the Lam´e parameters. They are the more simple energies coupling elasticity and volume damage. They give the constitutive laws

σ1 =

∂Ψ1 = β1 {λ1trε1 1 + 2μ1 ε1 }, ∂ ε1

∂Ψ1 ∂ Φ1 + ∂ β1 ∂ (∂ β1 /dt)      ∂ β1 ∂ β1 1 2 + ∂ I− , = −w1 + λ1 (trε1 ) + 2μ1 ε1 : ε1 + ∂ I(β1 ) + c 2 ∂t ∂t

B1 =

H1 =

∂Ψ1 = k1 grad β1 , ∂ (grad β1 )

where 1 is the identity matrix. The free energy and pseudopotential of the glued contact surface are

8.3 The Constitutive Laws

121

Ψs (Es ) = Ψs (u2 − u1 , βs , grads βs , β1 − βs , β2 − βs) = ws (1 − βs) +

ks (grads βs )2 + I(βs ) + I− ((u2 − u1 ) · N2 ) 2

ks,1 ks,2 βs kˆ s (u2 − u1 )2 + (β1 − βs )2 + ( β2 − βs ) 2 , 2 2 2     ∂ βs ∂ βs c s ∂ βs 2 Φs (δ Es ) = Φs ( + I− , )= ∂t 2 ∂t ∂t +

where ws is the surface damage threshold, ks the surface damage interaction parameter, kˆ s is the surface rigidity, ks,1 and ks,2 are the damage surface-volume interaction parameter, cs is the damage viscosity. The function I− ((u2 − u1 ) · N2 ) takes into account the impenetrability of the two pieces of concrete on their contact surface. The surface free energies and pseudopotential of dissipation are also the more simple we may choose. They account for elastic, viscous and damage properties. They give the constitutive laws

∂Ψs ∈ βs kˆ s (u2 − u1 ) + ∂ I− ((u2 − u1 ) · N2 ) N2 , ∂ (u2 − u1 )   kˆ s ∂Ψs ∂ βs ∂ βs ∈ −ws + (u2 − u1 )2 + ∂ I(βs ) + cs Bs = , + ∂ I− ∂ βs 2 ∂t ∂t R=

Hs =

(8.2)

∂Ψs = ks grads βs , ∂ (grads βs )

B1,s =

∂Ψs = ks,1 (β1 − βs), ∂ (β1 − β s )

B2,s =

∂Ψs = ks,2 (β2 − βs). ∂ (β2 − β s )

The force in ∂ I− ((u2 − u1 ) · N2 ) N2 is the impenetrability reaction. The non-local free energy on the glued contact surface is

Ψs,1,2 (Es,1,2 (x, y)) =

ks,1,2 2 |x − y|2 g (y, x) (βs (x)βs (y)) exp(− ), 2 d2

with g(y, x) = 2(y − x) · (u2 (y) − u1 (x)). The exponential function with distance d, describes the attenuation of non-local actions with distance |x − y| between points x and y. We assume no dissipation with respect to δ Es,1,2 (x, y) and have the constitutive law

122

8 Damage of Solids Glued on One Another

− Bs,1(x, y) =

ks,1,2 2 |x − y|2 ∂Ψs,1,2 ), (Es,1,2 (x, y)) = g (x, y)βs (y) exp(− ∂ βs (x) 2 d2

−Bs,2 (x, y) =

∂Ψs,1,2 ks,1,2 2 |x − y|2 (Es,1,2 (x, y)) = g (x, y)βs (x) exp(− ), ∂ βs (y) 2 d2

−M(x, y) =

∂Ψs,1,2 |x − y|2 (Es,1,2 (x, y)) = ks,1,2 g(x, y) (βs (x)βs (y)) exp(− ). ∂ g(y, x) d2

Let us note that all the constitutive laws involve the reactions to the internal constraints when needed, which are clearly non linear relationships, and linear relationships between the forces and the state quantities and velocities. Thus they are simple and we think that they have to account for the main physical phenomena: non linear constitutive laws are to be chosen only to make the results more precise and adapted to deal with a particular situation. But the linear relationships have to be sufficient to capture the main properties.

8.4 The Equations They result from the equations of motion and the constitutive laws. They are:

8.4.1 On the Contact Surface cs

  ∂ βs ∂ βs  − ks Δ s βs + ∂ I(βs ) + ∂ I− ∂t ∂t

ws − − ks

kˆ s (u2 − u1 )2 + ks,1 (β1 − βs) + ks,2 (β2 − βs ) 2

 ∂ Ω1 ∩∂ Ω 2

ks,1,2 2 |x − y|2 (g (y, x) + g2 (x, y)βs (y) exp(− )dy, in ∂ Ω 1 ∩ ∂ Ω 2 , 2 d2

∂ βs = 0, on ∂ (∂ Ω1 ∩ ∂ Ω2 ). ∂ ns

The last but one term is not 0 when u2 − u1 = 0. It is responsible for the damage resulting from elongation. The glue damage source in the right hand side results from the gap between the two solids, from the

8.4 The Equations

123

elongation (the non-local effect) and from the flux of damaging work coming from the concrete. It is proportional to the difference of damage between the concrete and the glue. Thus it is more difficult to damage the glue when the concrete is not damaged. In this case the glue cohesion is ws + ks,1 + ks,2 whereas it is ws when the concrete is completely damaged. The contact boundary conditions on the glued contact surface ∂ Ω1 ∩ ∂ Ω2 are ∀x ∈ ∂ Ω 1 ∩ ∂ Ω 2 , σ1 N1 (x) = βs kˆ s (u2 − u1 ) + ∂ I− ((u2 − u1 ) · N2 ) N2 −

 ∂ Ω 1 ∩∂ Ω 2

2(x − y)ks,1,2 g(x, y) (βs (x)βs (y)) exp(−

|x − y|2 )dΓ (y), d2

∀y ∈ ∂ Ω 1 ∩ ∂ Ω 2 , σ2 N2 (y) = −βs kˆ s (u2 − u1 ) − ∂ I− ((u2 − u1 ) · N2 ) N2 , − k1

 ∂ Ω 1 ∩∂ Ω 2

2(y − x)ks,1,2 g(x, y) (βs (x)βs (y)) exp(−

∂ β1 ∂ β2 = ks,1 (βs − β1 ), k1 = ks,2 (βs − β2). ∂ N1 ∂ N2

|x − y|2 )dΓ (x), d2 (8.3)

In the following numerical applications, we have neglected the non local mechanical effect on the contact surface stresses because it has not an important overall mechanical effect. The values of parameters kˆ s  ks,1,2 of the constitutive laws we choose in the sequel agree with this assumption. Let us recall that this non-local effect has been introduced to take into account damage which is produced by the elongation, i.e. by displacements such that u2 (x) − u1 (x) = 0, with u2 (x) − u1 (y) = 0, f or x = y. The boundary condition (8.3) means that the damaging work flux in the concrete is proportional to the difference of damage between the glue and the concrete.

8.4.2 In the Domains As already said, they are identical for the two domain. For domain Ω1 , they are div(β1 {λ1trε1 (u1 )1 + 2μ1ε1 (u1 )}) = 0,    1 ∂ β1 ∂ β1  w1 − − k1 Δ β1 + ∂ I(β1 ) + ∂ I− λ1 (trε1 )2 + 2μ1 ε1 : ε1 , c1 ∂t ∂t 2

124

8 Damage of Solids Glued on One Another

with initial conditions

β1 (x, 0) = β10 (x), in Ω1 , βs (x, 0) = β10 (x), on ∂ Ω 1 ∩ ∂ Ω 2 , and boundary conditions

σ1 N1 = g1 , on ∂ Ω1 \(∂ Ω1 ∩ ∂ Ω2 ), k1

∂ β1 = 0, on ∂ Ω1 \(∂ Ω1 ∩ ∂ Ω2 ). ∂ N1

8.5 Examples The following examples show how important are the interaction parameters ks,1 and ks,2 which couple the damages of solids 1 and 2: when solid 1 damages in the neighbourhood of solid 2, solid 2 damages also. The examples confirm also that it is more difficult to damage the glue when the concrete is not damaged than when the concrete is damaged: the glue cohesion or threshold is ws + ks,1 + ks,2 when the concrete is not damaged whereas it is ws when concrete is completely damaged in the two solids. The examples show also the important effect of the stretching described by the non local interactions. All the computations are due to Francesco Freddi, [10, 27, 28, 99, 100].

8.5.1 Four Points Bending Some experimental results due to Marie Paule Thaveau, [203], are reported on Fig. 8.1. One piece of concrete or two pieces of concrete glued on one another are tested. Their length is 0.4 m, their width is 0.3 m, their height is either 0.1 m or 2 × 0.05 m. The Young modulus is 38 G Pa, the Poisson modulus is 0.2. The maximum load is 14.3 kN for the one piece specimen and 18.2 kN for the two pieces specimen. Numerical results are shown on Figs. 8.2–8.5. They concern three models: 1. There is no damaging interaction between solids 1, 2 and the glue: ks,1 = ks,2 = 0. Stretching is not taken into account: ks,1,2 = 0. 2. There is damaging interaction between solids 1, 2 and the glue: ks,1 = 0, ks,2 = 0. Stretching is not taken into account: ks,1,2 = 0. 3. There is damaging interaction between solids 1, 2 and the glue: ks,1 = 0, ks,2 = 0. Stretching is taken into account: ks,1,2 = 0.

8.5 Examples

125

Fig. 8.1 Four points bending, experimental results due to Marie Paule Thaveau, [203]. The experiments are either with two pieces of concrete glued on one another (number 5) or with one piece of concrete (number 19)

Fig. 8.2 Load versus displacement curves for one piece and two pieces concrete specimens in four point bending. The short dashed line is for the one piece specimen. The red line, long dashed line and continuous line are for the two pieces concrete specimen and models 1, 2 and 3 respectively

126

8 Damage of Solids Glued on One Another

Fig. 8.3 The damage for the two pieces and the three models (model 1 at the top, model 2 in the middle, model 3 at the bottom). The applied displacement is u = 0.045 mm. The contact surface of model 1 is a barrier for damage. It is not very good. The model 2 allows interactions of the damages of the two pieces. The model 3 allows interactions of the damages and the effect of stretching. It seems to be the best model. The color scales are different in the pictures (the color scale is given on the right of them)

The first model is unable to account for experimental results: the contact surface is a barrier which stops the damage (see Fig. 8.4). The two other models are good but the best is the one which takes into account the stretching effect.

8.5.2 Pull Test A vertical force is applied to two glued pieces of concrete (Fig. 8.6). The relative stiffness of the concrete and of the glue governs the behaviour of the structure.

8.5 Examples

127

Fig. 8.4 The contact surface is a barrier in model 1 whereas the effect of damage interaction and the effect of stretching are important in the best model 3. The applied displacement is u = 0.039 mm

When glue is solid and the concrete is weak, damage occurs in concrete just under the contact surface (Fig. 8.7). If the glue is weak and the concrete is solid, separation of the two pieces occurs on the contact surface and the concrete is not damaged, (Fig. 8.8).

8.5.3 Fibre-Reinforced Polymers-Concrete Delamination The experiment is described in Fig. 8.9. Experimental and numerical results are shown in Fig. 8.10. It appears a thin damaged zone in the concrete as well as large displacements: they correspond to a layer of concrete which remains glued on the Fibre-reinforced polymers (FRP, [15–17, 59, 103]) in the experiments. The determination of the parameters of the predictive theory from practical engineering knowledge is described in [28] together with the practical computation of the maximal load which a structure can bear.

128

8 Damage of Solids Glued on One Another

Fig. 8.5 The sharp discontinuity of the horizontal displacement accounts for the fracture. Red is a positive displacement towards the right, green is a negative displacement towards the left

Fig. 8.6 Pull test with important contrasts between the glue and concrete stiffness properties

8.5 Examples

129

Fig. 8.7 The glue is much more solid than the concrete; The damage occurs within the concrete just under the contact surface whereas the glue does not break

Fig. 8.8 The glue is weaker than the concrete. Damage occurs only in the glue. The concrete even if it is somewhere red is not damaged or almost not damaged. Let recall that the color scales are unique for each picture

130

8 Damage of Solids Glued on One Another

Fig. 8.9 A pull experiment for a FRP reinforced structure

Fig. 8.10 Damage occurs in the concrete and not in the FRP in agreement with experiments. Note the very large displacement and the damaged concrete in the right up corner where the FRP is pulled

Remark 8.2. In collisions which are investigated in following chapters, the damage β may be discontinuous with respect to space jumping from β = 1, sound material, to β = 0, completely damaged material. When β becomes 0 on a line whereas it is 1 elsewhere, a fracture is created. Fractures resulting from collisions may be investigated with these tools. This point of view is developed in [100].

Chapter 9

Phase Change with Discontinuity of Temperature: Warm Water in Contact with Cold Ice

Let us consider warm water on cold ice, for instance water flowing on a frozen lake. On the surface of the lake, there is a discontinuity of temperature between the temperature T2 of the liquid water which is larger than the water freezing temperature T0 and the temperature T1 of the ice of the lake which is lower than T0 . It is to be known if the water freezes or if the ice thaws, [111, 153]. In order to answer this question, let consider water with temperature T2 > T0 in contact with ice with temperature T1 < T0 on a surface Γ (t) which is a free boundary. We intend to describe the evolution of the temperatures T1 and T2 and the motion of the free surface which separates water and ice. For the sake of simplicity, we assume the water and ice do not move and that their densities are the same, ρ .

9.1 Warm Water in Contact with Cold Ice Outside the free boundary, we assume the free energy and pseudopotential of dissipation (4.1) chosen for the Stefan problem. The basic equations of mechanics on the free boundary between water and ice have been given in Sect. 3.5. Let us choose the pseudopotential, [111, 153] as k Φ (m, [T ]) = IM (m) + [T ]2 , 2

(9.1)

where k is a positive constant and IM is the indicator function of the convex set M   M = m | minf ≤ m ≤ msup , with minf ≤ 0 ≤ msup . We have

Ψ (T, β ) = −CT ln T − β

L (T − T0) , T0

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 9, © Springer-Verlag Berlin Heidelberg 2012

131

132

9 Phase Change with Discontinuity of Temperature

[Ψ ] = − [CT ln T ] − s = C(1 +

L (T2 − T0 ) , T0

ln T1 + ln T2 L , )+ 2 2T0

and − [Ψ ] − s[T ] = [CT ln T ] − C(1 + = O([T ]2 ) +

L ln T1 + lnT2 ) [T ] + (T − T0 ) 2 T0

L (T − T0 ) , T0

because β1 = 0 and β2 = 1, where O([T ]2 ) is of order [T ]2 . Within the small perturbation assumption, we have − [Ψ ] − s [T ] =

L (T − T0 ) . T0

The constitutive laws (3.36) are − [Ψ ] − s [T ] ∈ ∂ IM (m) , which is within the small perturbation assumption L (T − T0) ∈ ∂ IM (m) , T0

(9.2)

Q = k [T ] .

(9.3)

and

The law (9.3) is a Fourier law: the entropy flux through the free boundary is proportional to the difference of temperature. Parameter k is the thermal conductivity between the cold and warm domains. The other law (9.2) describes the movement of the free boundary. The mass flow through the free boundary is bounded: it cannot be too large. The constitutive laws gives the difference between the average temperature and the phase change temperature versus the mass flow or versus the velocity of the free boundary: 1. If T − T0 < 0 then m = minf = −ρ W ≤ 0, the water freezes if the average temperature is low. 2. If T − T0 > 0 then m = msup = −ρ W ≥ 0, the ice melts if the average temperature is high.   3. If T − T0 = 0 then m ∈ minf , msup , depending on the sign of m, either the water freezes or the ice melts.

9.2 Mixture of Ice and Water in Contact with Cold Ice

133

This situation occurs when solid and liquid in contact have different temperatures. The partial differential equations describing the evolution result from the chosen free energy and pseudopotential of dissipation (4.1). They are C

∂ ln T1 − λ1 Δ ln T1 = 0 , ∂t

the entropy balance in the ice or the solid phase; C

∂ ln T 2 − λ2 Δ ln T2 = 0 , ∂t

the entropy balance in the water or the liquid phase; T1 + T2 − T0 ∈ ∂ IM (−ρ W ) , 2 1 ∂ ln T1 ∂ ln T2 (λ1 + λ2 ) = k(T2 − T1 ) , 2 ∂N ∂N ∂ T2 ∂ T1 − λ1 , −ρ LW = λ2 ∂N ∂N the two constitutive laws (3.36) and the energy balance (3.29) on the free boundary Γ (t), the contact surface between solid and liquid. The equations are completed by boundary conditions and the initial position of the free boundary, Γ (0). Remark 9.1. The predictive theory given above investigates the contact of two materials at different temperatures (in this case m = 0). The problem of warm rain falling on frozen ground or on ice has to be investigated by using both thermal equations and mechanical equations. The effect of the collision of the rain droplets with the ground has to be taken into account. Both the temperatures and the velocities are discontinuous. The above thermal equations have to be coupled with the equations of motion resulting from collision theory [112, 114]. This problem is investigated in a simplified version in the following chapter where solids colliding are points and in a complete but more sophisticated model in Chap. 11.

9.2 Mixture of Ice and Water in Contact with Cold Ice Let us consider a mixture of water and ice at temperature, T2 = T0 with 0 < β2 < 1, in contact with cold ice. Constitutive laws (9.2) becomes

β2 L (T1 − T0) ∈ ∂ IM (m) . T0

134

9 Phase Change with Discontinuity of Temperature

The equations of the predictive theory become C

∂ ln T1 − λ1 Δ ln T1 = 0 , ∂t

the entropy balance in the ice or the solid phase; L

∂β2 =0, ∂t

the entropy balance in the ice water mixture. In the mixture the temperature is the phase change temperature T0 ; T1 − T0 ∈ ∂ IM (−ρ W ) ,

λ1 ∂ ln T1 = k(T0 − T1 ) , 2 ∂N ∂ T1 −ρ Lβ 2W = −λ1 , ∂N

(9.4)

the two constitutive laws (3.36) and the energy balance (3.29) on the free boundary Γ (t), the contact surface between ice solid phase and mixture. The equations are completed by boundary conditions and the initial position of the free boundary, Γ (0). Because T1 − T0 < 0, we have −ρ W = minf . It results that W > 0 with W = W · N where normal vector N is directed from the ice domain with index 1 toward mixture domain with index 2 (see Sect. 3.5). Thus the mixture of water and ice freezes and the ice warms up as shown by relationship (9.4).

9.3 An Example: Rain Falling on a Frozen Ground Let consider rain falling on a frozen ground or on ice. We investigate the evolution of the water: does it freeze? Does it remain liquid producing a liquid layer on top of ice or frozen ground. The rain is considered as an homogeneous mixture of air and water with water mixture density ρ2 , temperature T2 and vertical rain velocity U2 ≤ 0 (the vertical direction is directed upward). The rain falls on the ice or on the possible unfrozen water layer due to the rain water already fallen. We assume that the ice and possible water layer have null velocity. The mechanical effects are neglected and only the thermal effects are taken into account. The rain mass water

9.3 An Example: Rain Falling on a Frozen Ground

135

flow is mrain =

−ρ1 ρ2U2 . ρ2 − ρ1

Note that it is negative (mass flow from side 1 toward side 2). The maximal flow of water which can be frozen when reaching the ground is −mrain . Thus we let mrain = minf . From experiments, we assume that the heat exchanges on the freezing line are proportional to the differences of temperature between the ice and the water. For the sake of simplicity, we also assume that the mass flow is bounded from above: m ≤ msup , with msup ≥ 0. From those assumptions, it results the pseudopotential of dissipation (9.1) is valid. The parameter k is the thermal conductivity of the ice-liquid water interface. A more sophisticated and realistic pseudopotential of dissipation can be chosen [112, 153]. The problem which is considered is to find the evolution of the free boundary: either it remains at the surface in contact with the air (the rain freezes completely) or it is covered progressively by a layer of water (the rain does not freeze). The constitutive law (9.2) is the criterion for the rain to freeze or not to freeze, see Fig. 9.1: 1. If T − T0 < 0, the rain freezes immediately when touching the ice or the frozen ground and m = mrain , giving the velocity of the freezing line W =−

mrain . ρ1

2. If T − T0 > 0, the rain does not freeze and the ice or the frozen ground thaws. The velocity of the thawing line is m = msup ,

Fig. 9.1 Relationship (9.2) means that m = msup if T − T0 > 0, m = minf if T − T0 < 0 and, m ∈ minf , msup , if T − T0 = 0

136

9 Phase Change with Discontinuity of Temperature

W =−

msup . ρ1

3. If T − T0 = 0, the velocity of the freezing or thawing line is such that   W ∈ −(msup /ρ1 ), −(mrain /ρ1 ) ,   because m ∈ mrain , msup . A part of the water freezes if W > 0 and the ice or the frozen ground thaws if W < 0.

9.3.1 A Dissipative Behaviour Let us choose the pseudopotential of dissipation, [111, 112, 153]  k k Φ (m, [T ]) = m2 + IC(m) + [T ]2 , 2 2 where  k is a positive constant and C is the convex set, C = {m | minf ≤ m} , with minf < 0. The constitutive law (9.3) is still valid. The new constitutive law is, see Fig. 9.2

Fig. 9.2 The graph ∂ IC(m) + km. If 

T − T0 ≤ kL T0 minf , the solution of (9.5) is m = minf . If  T − T0 ≥ kL T0 minf , the solution is m =

T0 (T −T0 ) kL

9.3 An Example: Rain Falling on a Frozen Ground

137

L (T − T0 ) ∈  km + IC(m) . T0

(9.5)

Let consider rain with temperature 1 ◦ C, density ρ2 = 0.1 kg /m3 , mass flow mrain = −2 × 10−4 kg /(m2 s) which falls on a 5 mm ice layer at −3 ◦ C on a frozen ground with thickness 1m. The values of the dissipative parameters are  k = 183 m/ s, minf = mrain = −2 × 10−4 kg /(m2 s), k = 3.66 W /(m2 K2 ). At the beginning the water freezes when falling on the ice. The ice-layer thickens and the heat provided by the rain warms up the ice. As long as the temperature of the ice is such that L (T − T0 ) ≤  kmrain , T0 the rain freezes, see Fig. 9.3. When the ice is warmed enough, i.e., when its temperature is such that L (T − T0 ) ≥  kmrain , T0 it thaws. The rain no longer freezes and a water layer forms on top of the ice layer which is thawing, see Fig. 9.3. This phenomenon begins at time t = 954 s. At that time the temperature of the ice is −1.016 ◦C and the thickness of the ice layer is 8 7 6 5 4 3 2 1 0 0

2000

4000

6000

8000

10000

12000

14000

Fig. 9.3 The position (mm) of the top of the structure versus time (s), the ice at the beginning (in blue), the water at the end (in red), versus time (s). The position of the free surface (in blue) which separates the ice which is thawing from the water layer made of rain and thawed ice

138

9 Phase Change with Discontinuity of Temperature 1.5 1 0.5 0 −0.5 −1 −1.5 −2 −2.5 −3

0

2000

4000

6000

8000

10000 12000 14000

Fig. 9.4 The temperature of the top of the ice layer in blue and the temperature of the bottom of the water layer in red versus time. The difference of temperature becomes rapidly almost 0 ◦ C if thermal conductivity k is large. In this example, it has been chosen small to emphasis the phenomena

5.12 mm. Then the temperatures of the ice and water in contact tend rapidly to 0 ◦ C, Fig. 9.4. Remark 9.2. When the rain freezes, the ice receives heat from the rain but it also receives heat from the surrounding air. Thus the equations have to be slightly modified to take into account this heat intake. The basic modification are on the energy balance and fundamental inequality which become, m [e] = [T Q] + T1 B1 + T2B2 , H = m [s] − [Q] − B ≥ 0, where T1 B1 is the heat received by the rain from the air, T2 B2 is the heat received by the ice from the air and, B = (B1 + B2 )/2. It is easy to take these quantities into account, [112, 153]. One can, for instance, assume that the heat received from the air by the ice is proportional to the difference of the temperatures of the air and ice T2 B2 = −α2 (T2 − Tair ) and, assume that the rain is in thermal equilibrium with the air T1 B1 = T1 Q2 . The results shown have been computed with this assumption (α2 = 10 W /(m2 K)). At time 600 s the rain is freezing. The depth versus temperature in the whole ice-ground structure is shown in Fig. 9.5. One can see the effect of the rain which warms up the top of the structure. Figure 9.6 shows depth versus temperature just after the beginning of the ice thaw (t = 1029 s). It is possible to see in red the very thin layer of water made of thawed

9.3 An Example: Rain Falling on a Frozen Ground

139

1.2

1

0.8

0.6

0.4

0.2

0 −3

−2.5

−2

−1.5

−1

−0.5

0

0.5

1

Fig. 9.5 At time 600 s, the depth in the ground (m) versus temperature. The rain warms up the ground 1.009 1.008 1.007 1.006 1.005 1.004 1.003 1.002 1.001 1 −3

−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

Fig. 9.6 The depth versus temperature in the ice layer and the thin water layer just after the rain has stopped to freeze. The discontinuity of temperature between the water and ice is decreasing with time

ice and rain. The discontinuity of temperature is still visible. Figure 9.7 shows the temperature inside the whole structure. After four hours of rain, the layer of ice is almost thawed, see Fig. 9.8. And the discontinuity of temperature is no longer noticeable.

140

9 Phase Change with Discontinuity of Temperature 1.2

1

0.8

0.6

0.4

0.2

0 −3

−2.5

−2

−1.5

−1

−0.5

0

0.5

1

1.5

Fig. 9.7 The depth in the ground (m) versus temperature when the rain no longer freezes. Note a thin layer of water on top of the ice layer 1.2

1

0.8

0.6

0.4

0.2

0 −2

−1.5

−1

−0.5

0

0.5

1

1.5

Fig. 9.8 The depth versus temperature in the structure after four hours. The ice layer is almost thawed. There is no longer a temperature discontinuity between ice and water

Remark 9.3. The value of the thermal conductivity k is crucial for the decrease of the temperature discontinuity, [153]. For k = 11 W /(m2 K2 ), the temperatures of the water and of the ice become almost immediately equal to 0 ◦ C as soon as the rain no longer freezes. This value is more realistic than the value chosen in the present example which has the advantage to emphasis what occurs.

Chapter 10

Phase Change and Collisions

Let us consider warm rain falling on a frozen ground. On the surface of the ground, there is a discontinuity of temperature between the temperature T2 of the rain which is larger than the water freezing temperature T0 and the temperature T1 of the frozen ground which is lower than T0 . It is to be known if the rain freezes or if the frozen ground thaws. In this problem, the temperatures are discontinuous both with respect to space and with respect to time. Let us also note that the velocities are also discontinuous. In order to investigate this problem, we recall the theory of collisions without thermal effects and the theory of collision with thermal effects. For the sake of simplicity, we consider first collisions of balls schematized by points then in next chapter, collisions of continuous media either solid or liquid. This theory is developed in [114].

10.1 Collisions of Two Balls: The Mechanical Theory Let us consider two balls moving on a line, Fig. 10.1. For the sake of simplicity, we assume they are points with mass mi . They have position xi (t) and velocity Ui (t) = (dxi /dt)(t). Because the points can collide, we suppose that the velocities can be discontinuous, assuming the collisions to be instantaneous. The velocity before a collision at time t is Ui− (t) and the velocity after is Ui+ (t). In terms of mathematics, it seems interesting to think that the velocities are bounded variation functions of time. In the sequel we will use virtual velocities: a set of virtual velocities V = (V1 ,V2 ) is such that the two functions t → Vi (t) are bounded variation functions. We denote S(U,t1 ,t2 ) the set of discontinuity times of velocity U = (U1 ,U2 ) between times t1 and t2 (t1 < t2 )    S(U,t1 ,t2 ) = tk ∈ ]t1 ,t2 [ U + (tk ) = U − (tk ) .

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 10, © Springer-Verlag Berlin Heidelberg 2012

141

142

10 Phase Change and Collisions

Fig. 10.1 Two balls schematized by points move on an axis (their mass moments of inertia are 0). The mass of the balls are m1 and m2 , they have velocities U1 and U2

This set is numerable because the two functions Ui are functions of bounded variation. We denote [Ui (tk )] = Ui+ (tk ) − Ui− (tk ), the velocity discontinuity and in general [A(tk )] = A+ (tk ) − A−(tk ) the discontinuity of function t → A(t). More precisely, we assume that velocities Ui are special bounded variation functions, [5, 18, 54, 172]: their differential  dUi =

dUi dt

 dt +

∑

[Ui (tk )] δ (t − tk ),

tk ∈S(U,t1 ,t2 )

is the sum of a Lebesgue measure, the smooth part of the differential, whose density is {dUi /dt} and of a Dirac measure, the non-smooth part of the differential, whose density at point tk is discontinuity [Ui (tk )]. Remark 10.1. Velocity U + (t) depends on the future. It intervenes in the theory, particularly in the constitutive laws. This seems to contradict the causality principle. This is not the case, as already mentioned in Sect. 3.5. Velocity U + (t) sums up the sophisticated evolution taking place during the physical phenomenon responsible for the velocity discontinuity. Thus it depends on the past.

10.1.1 The Velocities of Deformation It is easy to write the equations of motion for the two points. But in order to be more precise and have results which can be easily adapted to more general settings, we derive carefully these equations by using the principle of virtual work. Before we introduce it, let us remark that there is no reason to speak of deformation when dealing with an isolated point. But if we consider the system made of the two points, its form changes because the distance of the two points may change. Thus it is wise to consider that the system is deformable. A way to characterize the system velocity of deformation is to consider the relative velocity of the two points D(U) = U1 − U2,

10.1 Collisions of Two Balls: The Mechanical Theory

143

where U = (U1 ,U2 ) is the set of the two actual velocities of the points. A rigid system set of velocities is such that the form of the system does not change: it is easy to see that they are characterized by D(U) = U1 − U2 = 0, because when the form of the system does not change, the velocities of the two points are equal.

10.1.2 The Principle of Virtual Work The principle is The virtual work of the acceleration forces is equal to the sum of the virtual work of the interior forces and the virtual work of the exterior forces. More precisely, for any time t1 , any time t2 > t1 and any virtual velocity V Tacc (t1 ,t2 ,U,V ) = Tint (t1 ,t2 ,U,V ) + Text (t1 ,t2 ,U,V ), where Tacc (t1 ,t2 ,U,V ) is the virtual work of the acceleration forces between times t1 and t2 , Tint (t1 ,t2 ,U,V ) and Text (t1 ,t2 ,U,V ) are the virtual works of the interior and exterior forces between the same times. These works satisfy: 1. The virtual work of the acceleration forces is a linear function of the virtual velocity V , when S(V,t1 ,t2 ) is fixed. The actual work of the acceleration forces is equal to the variation of the kinetic energy between times t1 and t2 Tacc (t1 ,t2 ,U,U) = K (U − (t2 )) − K (U + (t1 )), where K is the system kinetic energy. This property and the principle of virtual work with the actual velocities, i.e., V = U, are sometimes called the theorem of expanded energy (teorema dell’energia cinetica in Italian and le th´eor`eme de l’´energie cin´etique in French). 2. The virtual work of the interior forces is a linear function of the virtual velocity V , when S(V,t1 ,t2 ) is fixed. It satisfies Tint (t1 ,t2 ,U,V ) = 0, for any rigid system velocity. This relationship is often called the axiom of the principle of virtual work, [132, 197]. 3. The virtual work of the exterior forces is a linear function of the virtual velocity V , when S(V,t1 ,t2 ) is fixed. Remark 10.2. The principle of virtual work we use here is not to be confused with the principle of virtual power where the velocities are understood as small

144

10 Phase Change and Collisions

displacements (This relationship is often also called misleadingly the principle of virtual work).

10.1.3 The Virtual Works Now we may introduce interior forces to the deformable system. A productive way is to define these generalized forces by their work (or by duality in terms of mathematics). The work of the interior forces of the system between times t1 and t2 (t1 < t2 ) is a linear function of the velocity of deformation. We choose as virtual work of the interior forces Tint (t1 ,t2 ,U,V ) = −

 t2 t1

Rint (τ )D(V )(τ )d τ

∑

−

tk ∈S(U,t1 ,t2 )∪S(V,t1 ,t2

   + V (tk ) + V − (tk ) Pint (tk )D , 2 )

which is a linear function of D(V ) = (V1 −V2 ) when S(V,t1 ,t2 ) is fixed. Sign minus has no importance, it is chosen in accordance with habits of continuum mechanics. Its advantage will appear down below where a classical inequality has its classical sign. The virtual work defines an interior force Rint (τ ) which intervene in the smooth evolution when the two points evolve without colliding (Rint (τ ) = 0 when the two points do not interact at a distance). It defines also an interior percussion Pint (t) which intervene when collisions occur. Remark 10.3. The linear function of D(V ) we have chosen is not the more general. The choice involving −Pint+ (tk )D(V + )(tk ) − Pint− (tk )D(V − )(tk ), does not give much more opportunities, [112]. Let us note that Tint (t1 ,t2 ,U,V ) = 0 for any rigid system velocity, i.e., for D(V ) = (V1 − V2) = 0. The virtual work of the acceleration forces is Tacc (t1 ,t2 ,U,V ) =

 dUi ( τ )V ( τ )d τ i ∑ t dt 1 i=1

  2 Vi+ (tk ) + Vi− (tk ) +∑ mi [Ui (tk )]. , ∑ 2 i=1 t ∈S(U,t ,t )∪S(V,t ,t ) 2



t2

k

mi

1 2

1 2

where [U(tk )] = U + (tk ) − U − (tk ) is the velocity discontinuity. It is clear that this work is a linear function of V when S(V,t1 ,t2 ) is fixed.

10.1 Collisions of Two Balls: The Mechanical Theory

145

10.1.3.1 The Theorem of Expanded Energy Let us compute the actual work of the acceleration forces Tacc (t1 ,t2 ,U,U)

   2 t2 Ui+ (tk ) + Ui−(tk ) dUi mi [Ui (tk )] . (τ )Ui (τ )d τ + =∑ mi ∑ dt 2 t1 i=1 tk ∈S(U,t1 ,t2 )

 

2 t2 dUi mi 2 mi =∑ (τ )Ui (τ )d τ + ∑ 2 (Ui (tk )) dt t1 i=1 t ∈S(U,t ,t ) 1 2

k

=

m  2 mi  + 2  i = K (U − (t2 )) − K (U + (t1 )) Ui− (t2 ) − Ui (t1 ) 2 2 i=1 2

∑

= K − (t2 ) − K + (t1 ), which is equal to the variation of kinetic energy K between times t1 and t2 . At any time we have kinetic energies K − and K + in case there is a collision. The virtual work of the exterior forces is a linear function of the virtual velocities V when S(V,t1 ,t2 ) is fixed, we choose

=

Text (t1 ,t2 ,U,V )

 2

∑

i=1

t2

t1

∑

Fi (τ )Vi (τ )d τ +

tk ∈S(U,t1 ,t2 )∪S(V,t1 ,t2

  + − V (t ) + V (t ) k k i Piext (tk ) i . 2 )

Exterior force Fi may be the gravity force. Exterior percussion Piext may represent hammer blows applied to the points. Remark 10.4. It is not common that an exterior percussion is applied to a mechanical system at the same time a collision occurs. An example is in a pin-ball machine where a collision may produce an external electrical blow.

10.1.4 The Equations of Motion We assume that the interval ]t1 ,t2 [ is split in distinct intervals ]tk ,tl [ where there is no discontinuity of U, the two ends of the intervals being discontinuity times. By choosing smooth virtual velocities with compact support in interval ]tk ,tl [, we get 2

∑

 tl

i=1 tk

mi

dUi (τ )Vi (τ )d τ = − dt

 tl tk

2

Rint (τ )D(V )(τ )d τ + ∑

 tl

i=1 tk

Fi (τ )Vi (τ )d τ .

146

10 Phase Change and Collisions

The fundamental lemma of variation calculus gives dU1 = −Rint + F1, a.e. in ]tk ,tl [ , dt dU2 = Rint + F2 , a.e. in ]tk ,tl [ . m2 dt m1

(10.1)

Remark 10.5. a.e.means almost everywhere or almost always in this context. Because points ti are numerable, relationships (10.1) are satisfied almost everywhere in ]t1 ,t2 [. Thus the principle becomes   Vi+ (tk ) + Vi− (tk ) [U (t )] m i i k ∑ ∑ 2 i=1 tk ∈S(U,t1 ,t2 )∪S(V,t1 ,t2 )  +   V (tk ) + V − (tk ) int P (tk )D =− ∑ 2 t ∈S(U,t ,t )∪S(V,t ,t ) 2

k

1 2

2

+∑

1 2

∑

i=1 tk ∈S(U,t1 ,t2 )∪S(V,t1 ,t2 )

 Piext (tk )

 Vi+ (tk ) + Vi− (tk ) . 2

At time t j between intervals ]tk ,t j [ and ]t j ,tl [, we have by choosing virtual velocities with compact support in ]tk ,tl [ Vi+ (t j ) + Vi− (t j ) 2 i=1 2  + V + (t j ) + Vi− (t j ) V (tk ) + V − (tk ) + ∑ Piext (t j ) i , = −Pint (t j )D 2 2 i=1 2

∑ mi [Ui (t j )]

which gives immediately m1 [U1 (t j )] = −Pint (t j ) + P1ext (t j ),

(10.2)

m2 [U2 (t j )] = Pint (t j ) + P2ext (t j ).

(10.3)

At time t ∈ ]tk ,tl [ by choosing virtual velocities with support in ]tk ,tl [ having a unique discontinuity point at time t, we get easily that relationships (10.2) and (10.3) are satisfied at time t. Thus, they are satisfied at any time m1 [U1 (t j )] = −Pint (t j ) + P1ext (t j ), ∀t ∈ ]t1 ,t2 [ , m2 [U2 (t j )] = Pint (t j ) + P2ext (t j ), ∀t ∈ ]t1 ,t2 [ . A detailed derivation of the equations of motion is given in [113].

10.1 Collisions of Two Balls: The Mechanical Theory

147

Remark 10.6. The equations of motion are a simplified version of the equations given in Sect. 3.6 for two solids when the velocities and temperatures are discontinuous on the whole domains occupied by the two solids at time of collision.

10.1.5 An Example of Collisions: The Equations for U + The constitutive laws, given in a general setting in next section, define the interior percussion Pint with a pseudopotential of dissipation Φ depending on D = D((U + + U − )/2) and on parameter χ = D(U − /2) which depends on the past. The noninterpenetration condition of the two balls is an internal constraint U1+ − U2+ = D(U + ) ≤ 0. It is taken into account by the pseudopotential of dissipation

Φ (D, χ ) = kD2 + I− (D − χ ),

(10.4)

where I− is the indicator function of the set of the non positive real numbers. In order to solve the equations of motion, we rewrite them in a way adapted to numerical computation. We denote ˆ D(U) = EU, with deformation matrix E

  E = 1 −1 ,

and

  U1 Uˆ = . U2

Constitutive law is   +  − U +U− U Pint = Pd + Preac ∈ ∂ Φ D ,D , 2 2  + U +U− Pd = 2kD = kD(U + + U −), 2  + U− U +U− reac − D( )) = ∂ I− (U1+ − U2+), P ∈ ∂ I− (D 2 2

(10.5)

where subdifferential ∂ Φ of Φ (D, χ ) is computed with respect to D, see Appendix A. Equations of motion (10.2) and (10.3) are M [U] + E T Pint = 0,

(10.6)

148

10 Phase Change and Collisions

where matrix E T is the transposed matrix of E and the mass matrix is 

 m1 0 M= . 0 m2 Equations (10.5) and (10.6) give   +  − U +U− U M [U] + E ∂ Φ D ,D  0, 2 2 T

or

  +  − U +U− U ,E  0. M [U] + E ∂ Φ E 2 2 T

By letting X = U + + U − , the equation becomes    − U X ,E  2MU − . MX + E T ∂ Φ E 2 2

(10.7)

Note that to find U + when U − is known, is equivalent to find X. Equation (10.7) has good properties: Theorem 10.1. If dissipation coefficient k is positive, (10.7) has one and only one solution. Proof. Operator

   − Y U Y → ET ∂ Φ E ,E , 2 2

is monotone and maximal from R2 into R2 because Φ is convex. Operator    − Y U ,E , Y → MY + E ∂ Φ E 2 2 T

is strictly monotone, maximal and surjective from R2 onto R2 . It results (10.7) has one and only one solution.

This result is general. It applies to collisions of a system involving any number of rigid bodies, [112]. The equations are used to perform numerical computations for engineering purposes, [83]. 10.1.5.1 Computation of the Velocities U + : Computation of the Dissipated Work Let us compute the velocities U1+ and U2+ after a collision assuming the velocities before the collision U1− and U2− , which satisfy U1− −U2− > 0, are known. Depending

10.1 Collisions of Two Balls: The Mechanical Theory

149

on the values of the mi and k, there are two possible evolutions. The equations giving U1+ , U2+ depending on U1− , U2− and satisfying the impenetrability condition U1− − U2− ≥ 0, are m1 (U1+ − U1− ) = −Pint = −k(U1+ − U2+ + U1− − U2− ) − Preac , Preac ∈ ∂ I− (U1+ − U2+ ), m2 (U2+ − U2− ) = Pint , where Preac is the impenetrability reaction (∂ I− is the subdifferential set of function I− (see Appendix A), Preac is 0 if contact is not maintained after collision (U1+ − U2+ < 0), it is non negative if contact is maintained after collision (U1+ − U2+ = 0). We get by solving the previous equations or (10.7): 1. If m1 m2 − k(m1 + m2 ) ≤ 0, i.e., if the balls are not heavy or if system is very dissipative, they rebound with velocities U1+ =

(m1 m2 + km1 − km2 )U1− + 2km2U2− , m1 m2 + km1 + km2

U2+ =

(m1 m2 + km2 − km1 )U2− + 2km1U1− , m1 m2 + km1 + km2

(10.8)

Preac = 0, U1+ − U2+ ≤ 0.  + − which intervene in the thermal pheThe dissipated work Pint D U +U 2 nomenon described in next section is  + k U +U− int P D = (U1+ − U2+ + U1− − U2−)2 2 2 =

2km21 m22 (m1 m2 + km1 + km2 )

2

 − 2 U1 − U2− .

2. If m1 m2 − k(m1 + m2 ) ≥ 0, i.e., if the balls are heavy or if system is not very dissipative, they do not rebound. Their common velocity is U1+ = U2+ =

m1U1− + m2U2− . m1 + m2

The impenetrability reaction is Preac =

m1 m2 − k(m1 + m2 ) − (U1 − U2−) ≥ 0, m1 + m 2

because U1+ − U2+ = 0. The dissipated work is

150

10 Phase Change and Collisions



U+ +U− P D 2 int



U − − U2− k + (U1 − U2+ + U1− − U2− )2 + Preac 1 2 2 m 1 m2 = (U − − U2−)2 . 2(m1 + m2 ) 1

=

3. If m1 m2 − k(m1 + m2 ) = 0, case 1 and 2 solutions are valid with U1+ − U2+ = 0, Preac = 0. 10.1.5.2 The Measure of Dissipation Parameter k In case the two balls rebound, we have U1+ − U2+ =

 m1 m2 − k(m1 + m2 )  − U1 − U2− . m1 m2 + km1 + km2

Experiments give easily coefficient k by plotting the relative velocity after collision versus the relative velocity before. In case the balls do not rebound, positive exterior percussion P2ext may be applied at ball number 2 and negative exterior percussion P1ext may be applied at ball 1, concomitant to collision, in such a way that U1− +

P1ext Pext − (U2− + 2 ) < 0, m1 m2

which gives U1+ = U2+ and U1+ − U2+ =

Pext m1 m2 − k(m1 + m2 ) − P1ext (U1 + − (U2− + 2 )), m1 m2 + km1 + km2 m1 m2

allowing to measure k. In practice, experiments may be performed with the balls at rest before collision. In this situation, parameter k quantifies the difficulty to separate the balls in contact. It accounts for adhesion effects. The larger it is, the more difficult is the separation. Other way to describe adhesion are given in [83]. This example, shows how to identify more sophisticated constitutive laws by  assuming that k depends on D = U1+ − U2+ + U1− − U2− /2 and χ = U1− − U2− .

10.1.6 An Other Example: The Impenetrability Condition The impenetrability condition is a physical property. There are case when it is conditional: let consider for instance, the solid number 1 to be a steel ball and solid number 2 to be a paper sheet which is hold by somebody. We let the steel

10.1 Collisions of Two Balls: The Mechanical Theory

151

ball fall on the sheet of paper (on Fig. 10.1, the downward vertical direction is toward the right). When the steel ball collides the sheet of paper, depending on its weight and velocity, it goes through the sheet or it is stopped. In this situation, the impenetrability depends on the circumstances. For the sake of simplicity, we assume that the phenomenon is unidimensional and that the sheet of paper is firmly hold in such a way its velocity remains constant, U2− = U2+ = 0. When colliding, the velocity of ball number 1 is U1− > 0. We choose the pseudopotential of dissipation describing the collision properties of the system (steel ball)∪(sheet of paper) to be

Φ (D, χ ) = kD2 + k0 pp (D − χ ), where pp(x) = sup(x, 0) is the positive part of x and D=

U+ +U− U− ,χ= , 2 2

denoting U = U1 . Dissipative parameters k and k0 are positive. The function D → Φ (D, χ ) is a pseudopotential of dissipation because

Φ (0, χ ) = Φ (0,

U− U− ) = pp(− ) = 0, 2 2

because U − > 0. The interior percussion is  P

int

∈ ∂Φ

where the graph H is

U+ +U− 2



    = k U + + U − + k0 H U + ,

⎧ ⎨ 1, H(x) = [0, 1], ⎩ 0,

if x > 0, if x = 0, if x < 0.

The equation to find the velocity of the steel ball U + after the collision with the sheet of paper is m(U + − U − ) = −Pint ∈ −k(U + + U −) − k0 H(U + ). By letting X = U + + U − , it becomes mX + k0 H(X − UN−) + kX  2mUN− . Applying a theorem analog to Theorem 10.1, the preceding equation gives one and only one velocity U + when velocity U − is known. We have:

152

10 Phase Change and Collisions

Fig. 10.2 The velocity U + of the steel ball versus its velocity U − before colliding the sheet of paper. When the falling velocity is low, U − ≤ k0 / (m − k), the ball is stopped by the sheet of paper. When it is large, U − ≥ k0 / (m − k), the steel ball goes through the sheet of paper

1. If m − k ≤ 0, i.e., if the steel ball is very light or the system (steel ball)∪(sheet of paper) is very dissipative, the ball rebounds with velocity U+ =

m−k − U . m+k

2. If m − k ≥ 0, i.e., if the steel ball is heavy or the system is not very dissipative, we get, Fig. 10.2: a. If the falling velocity is small U− ≤

k0 , m−k

the steel ball does not bounce. It is stopped by the sheet of paper U + = 0. b. If the falling velocity is large U− ≥

k0 , m−k

the steel ball goes through the paper sheet U+ =

(m − k)U − − k0 , m+k

U − > U + ≥ 0. After collision, the ball is still falling but it has been slowed down when going through the paper sheet.

10.2 Collisions of Two Balls: The Thermal Phenomenon

153

In Case 2 the dissipated work is Pint

U+ +U− m − 2 = (U ) , 2 2

if the falling velocity is slow, and int U

P

+ +U−

2

m = 2



4km(U − )2 + 2k0 (m − k)U − − k02 (m + k)2

 ,

if the falling velocity is high. This example is due Eric Dimnet in order to describe rock avalanches in mountain forests. The rocks may break trees and continue their way down the slope, [84]. Remark 10.7. In both examples, the dissipated work increases with the relative velocity U1− − U2− or U − .

10.2 Collisions of Two Balls: The Thermal Phenomenon Four temperatures appear in collisions : Ti+ and Ti− temperatures of the points after and before collision. We denote [T1 ] = T1+ − T1− , [T2 ] = T2+ − T2− . These differences are analog to the temperature time derivative dT /dt when the temperature is smooth. Let us recall that there is no dissipation with respect to this quantity because of Helmholtz relationship (see formula (3.8)) s=−

∂Ψ (T, η ), ∂T

defining the entropy s when E = (T, η ) (see Sect. 3.2.3). We denote some average temperatures T¯1 =

T1+ + T1− T + + T2− ¯ T + + T1− + T2+ + T2− T¯1 + T¯2 , T¯2 = 2 ,Θ= 1 = . 2 2 4 2

The average temperature difference

δ T¯ = T¯2 − T¯1, is the analog of the spatial thermal gradient in a smooth situation. Let us also recall that there is dissipation with respect to this quantity: the almost universal example

154

10 Phase Change and Collisions

is the Fourier law. In the following investigation of the thermal effects of collisions, analogous choices are made remembering that they sum up what occur during the duration of collisions when Fourier law is assumed and Helmholtz relationship is valid. In the sequel, we consider sums which are transformed in the following way T + B+ + T − B − = T¯ Σ (B) + [T ] Δ (B), with T+ + T− B+ − B− , [T ] = T + − T − , Σ (B) = B+ + B − , Δ (B) = . T¯ = 2 2

10.2.1 First Law of Thermodynamics for a Point The laws of thermodynamics are the same for the two points. For point 1, the first law is [E1 ] + [K1 ] = Text1 (U) + C1 , where Text1 (U) is the actual work of the percussions which are exterior to point 1 and C1 is the amount of heat received by the point in the collision. This heat involves the heats T1 B1 received from the exterior to the system and the heats T1 B12 received from the interior of the system, i.e., from the other point. We assume that these heats are received either at temperature T1+ or at temperature T1−     + − C1 = T1+ B1+ + B12 + T1− B1− + B12 = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 )). The theorem of expanded energy for point 1 is [K1 ] = Tacc1 (U) = Text1 (U). It gives with the energy balance [E1 ] = C1 = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 )).

(10.9)

In the right hand side, there is no mechanical quantity because a point has no interior force, the work of which could contribute to the internal energy evolution.

10.2.2 Second Law of Thermodynamics for a Point It is

+ − [S1 ] ≥ B1+ + B1− + B12 + B12 = Σ (B1 ) + Σ (B12 ).

10.2 Collisions of Two Balls: The Thermal Phenomenon

155

10.2.3 A Useful Inequality for a Point Second law and relationship (10.9) give [E1 ] − T¯1 [S1 ] ≤ [T1 ] (Δ (B1 ) + Δ (B12 )), or by introducing Ψ = E − T S [Ψ1 ] + S 1 [T1 ] ≤ [T1 ] Δ (B1 ) + [T1 ] Δ (B12 ). We denote

[Ψ1 ] + S 1 [T1 ] = −S1 [T1 ],

remembering that when [T1 ] → 0 we have [Ψ1 ] / [T1 ] + S 1 → 0 because ∂Ψ /∂ T + s = 0. Then 0 ≤ [T1 ] (Δ (B1 ) + Δ (B12 ) + S1 ). Due to Helmholtz relationship, it is reasonable to assume no dissipation with respect to [T1 ] which is, as already said, the analog to dT /dt

Δ (B1 ) + Δ (B12 ) + S1 = 0. We deduce [S1 ] = Σ (B1 ) + Σ (B12 ).

(10.10)

10.2.4 The First Law for the System The internal energy and entropy of the system are the sums of the internal energies and entropies of its components E1 , S1 and E2 , S2 to which interaction internal energy and entropy E int , S int are added: E = E1 + E2 + E int , S = S1 + S2 + S int . The first law for the system is [E ] + [K ] = Text (U) + C , where C is the heat received by the system in collision C = T1+ B1+ + T1− B1− + T2+ B2+ + T2− B2− .

156

10 Phase Change and Collisions

The theorem of expanded energy is [K ] = Tacc (U) = Tint (D(U)) + Text (U). It gives with the first law [E ] = −Tint (D(U)) + C .

(10.11)

The first laws for each point (10.9) and for the system (10.11) give   [E ] = [E1 ] + [E2 ] + E int   = E int + T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 )) +T¯2 (Σ (B2 ) + Σ (B21 )) + [T2 ] (Δ (B2 ) + Δ (B21 ))  + U +U− + T1+ B1+ + T1− B1− + T2+ B2+ + T2− B2− = Pint D 2  + U +U− int + T¯1 Σ (B1 ) + [T1 ] Δ (B1 ) + T¯2 Σ (B2 ) + [T2 ] Δ (B2 ). =P D 2 Then 

 +  U +U− E int = Pint D 2 −T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) − T¯2 Σ (B21 ) − [T2 ] Δ (B21 ). (10.12)

This relationship is interesting because only interior quantities intervene.

10.2.5 The Second Law for the System It is where

[S ] ≥ B1+ + B1− + B2+ + B2− = Σ (B1 ) + Σ (B2 ) ,   [S ] = [S1 ] + [S2 ] + S int ≥ Σ (B1 ) + Σ (B2 ) ,

which gives with (10.10)  int  ≥ −Σ (B12 ) − Σ (B21 ). S

(10.13)

In this relationship only interior heat exchanges intervene. Together with the first law (10.12), it gives an inequality coupling the mechanical and thermal dissipations.

10.2 Collisions of Two Balls: The Thermal Phenomenon

157

10.2.6 A Useful Inequality for the System We have with (10.12) and (10.13) 

 +    U +U− − T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) E int − Θ¯ S int ≤ Pint D 2 −T¯2 Σ (B21 ) − [T2 ] Δ (B21 ) + Θ¯ (Σ (B12 ) + Σ (B21 )).

We define the interaction free energy: Ψ int = E int − Θ S int . It gives with the preceding relationship  +  int  U +U− + S int [Θ ] ≤ Pint D Ψ 2 −T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) − T¯2 Σ (B21 ) − [T2 ] Δ (B21 ) +Θ¯ (Σ (B12 ) + Σ (B21 )), with S int =

S int+ + S int− . 2

Remembering that we have denoted [Θ ] = we define

1 ([T1 ] + [T2 ]) , 2

 int  Ψ + S int [Θ ] = −Sint [Θ ].

We obtain  0 ≤ Pint D

U+ +U− 2

− T¯1 Σ (B12 ) − [T1 ] (Δ (B12 ) −

−T¯2 Σ (B21 ) − [T2 ] (Δ (B21 ) −

Sint ) 2

Sint ) + Θ¯ (Σ (B12 ) + Σ (B21 )). 2

It is reasonable not to have dissipation with respect to [T1 ] and [T2 ] which are temperature variations with respect to time and not with respect to space

Δ (B12 ) −

Sint Sint = 0, Δ (B21 ) − = 0. 2 2

(10.14)

Then  0 ≤ Pint D

U+ +U− 2

− T¯1 Σ (B12 ) − T¯2 Σ (B21 ) + Θ¯ (Σ (B12 ) + Σ (B21 )) ,

158

10 Phase Change and Collisions

but

Σ (B21 ) − Σ (B12 ) T¯1 Σ (B12 ) + T¯2 Σ (B21 ) = Θ¯ (Σ (B12 ) + Σ (B21 )) + δ T¯ , 2 recalling that δ T¯ = T¯2 − T¯1 . Thus we get  0 ≤ Pint D

U+ +U− 2

− δ T¯

Σ (B21 ) − Σ (B12 ) . 2

(10.15)

This relationship is important due to the quantities which intervene : the one which characterizes the mechanical evolution D ((U + + U −)/2), and the other one which characterizes the spatial thermal heterogeneity δ T¯ . Let us stress its analogy with the spatial thermal gradient in smooth situation. Relationship (10.15) is a guide to choose the quantities which are to be related to the interior forces. When dealing with experimental results relationship (10.15) is useful to choose the quantities which are to be plotted ones versus the others: Pint versus D (U + + U − ) and Σ (B21 ) − Σ (B12 ) versus δ T¯ . Let us recall that on our way we have assumed no dissipation with respect to the temperature time variation.

10.2.7 The Constitutive Laws Quantities which describe the mechanical evolution and thermal heterogeneity are D ((U + + U − )/2) and δ T¯ . We define the constitutive laws with a pseudopotential of dissipation   + U +U− , δ T¯ , χ . Φˆ D 2 The quantity χ = D (U − /2) which depends on the past is used to take into account the non-interpenetration of the points. Then the constitutive laws are   +  −  U +U− U int Σ (B21 ) − Σ (B12 ) ˆ ¯ P , ∈ ∂Φ D ,δT,D , 2 2 2 where the subdifferential set is computed with respect to the two first variables. It is easy to show that inequality (10.15) is satisfied. The choice of the pseudopotential of dissipation is to be guided by experiments.

10.2 Collisions of Two Balls: The Thermal Phenomenon

159

10.2.8 An Example of Thermal Effects Due to Collisions We choose a pseudopotential which does not couple the two first variables insuring that the mechanical problem is decoupled from the thermal one, for instance  −   +   +  − U U +U− U +U− U ˆ ˆ ¯ Φ D ,δT,D = Φmeca D ,D 2 2 2 2 +

λ 2 (δ T¯ ) . 4

The pseudopotential Φˆ meca gives the percussion Pint . The mechanical problem is not investigated in this Section but two examples have been given in Sect. 10.1. Complete investigation and results may be found in [113]. The thermal constitutive law is Σ (B21 ) − Σ (B12 ) = −λ δ T¯ . We choose the same simple free energy for the points Ψ = −CT ln T and a zero interaction free energy, Ψ int = 0. This choice gives Sint = 0, and with (10.14)

Δ (B12 ) = 0, Δ (B21 ) = 0.

Let us assume that the mechanical problem is solved, i.e., the work Pint D((U + + U − )/2) is known. The thermal equations are [S1 ] = C ln

T1+ = Σ (B1 ) + Σ (B12 ) , T1−

[S2 ] = C ln

T2+ = Σ (B2 ) + Σ (B21 ) , T2−

and [E1 ] = C [T1 ] = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] Δ (B1 ) , [E2 ] = C [T2 ] = T¯2 (Σ (B2 ) + Σ (B21 )) + [T2 ] Δ (B2 ) , 

0= E

int





U+ +U− =P D 2 int

− T¯1 Σ (B12 ) − T¯2 Σ (B21 ).

These equations are coupled with the description of the thermal exchanges with the exterior of the system. For instance, collision may be either isentropic

Σ (B1 ) = Σ (B2 ) = 0,

160

10 Phase Change and Collisions

or adiabatic T1+ B1+ + T1− B1− = T¯1 Σ (B1 ) + [T1 ] Δ (B1 ) = 0, T¯2 Σ (B2 ) + [T2 ] Δ (B2 ) = 0. The solution of these equations gives the temperatures Ti+ and the different entropy fluxes Bi and Bi j . 10.2.8.1 The Isentropic Case The equations give C ln

T2+ T1+ = Σ (B ) , C ln = Σ (B21 ) , 12 T1− T2−

C [T1 ] = T¯1 Σ (B12 ) + [T1 ] Δ (B1 ) , C [T2 ] = T¯2 Σ (B21 ) + [T2 ] Δ (B2 ) , U+ +U− Pint D = T¯1 Σ (B12 ) + T¯2 Σ (B21 ) , 2 

Σ (B21 ) − Σ (B12 ) = −λ δ T¯ . Then we have C [T1 ] = T¯1 Σ (B12 ) + [T1 ] Δ (B1 ) , C [T2 ] = T¯2 Σ (B21 ) + [T2 ] Δ (B2 ) , T+ T+ U+ +U− = T¯1C ln 1− + T¯2C ln 2− , Pint D 2 T1 T2 

C ln

T1+ T2+ ¯ ¯ − + λ T2 = C ln − + λ T1. T2 T1

The two last equations give the temperatures T1+ and T2+ after a collision. The first two equations give the thermal exchanges with the exterior : they are computed with Δ (B1 ) and Δ (B2 ). 10.2.8.2 The Adiabatic Case The equations give [S1 ] = C ln

T1+ = Σ (B1 ) + Σ (B12 ) , T1−

[S2 ] = C ln

T2+ = Σ (B2 ) + Σ (B21 ) , T2−

10.2 Collisions of Two Balls: The Thermal Phenomenon

and

161

C [T1 ] = T¯1 Σ (B12 ) , C [T2 ] = T¯2 Σ (B21 ) ,

and  Pint D

U+ +U− 2

= C [T1 ] + C [T2 ],

[T2 ] [T1 ] C ¯ − C ¯ = −λ δ T¯ . T2 T1 The two last equations give the temperatures T1+ and T2+ after a collision. The first two give the entropic exchanges Σ (B1 ) and Σ (B2 ) with the exterior.

10.2.8.3 A Computation Computations of temperatures T1+ and T2+ in the two preceding situations may be performed with numerics or with closed form solutions which are sophisticated. A case is easily solved: we assume that the collision occurs at the ambient temperature and that the temperature variations are negligible compared with the ambient temperature. We may say that we assume small perturbations. This is the case for an every day experiment. In the isentropic and adiabatic case, we have ln

T+ [T ] [T ] [T ] −, ¯ −. T− T T T

In both cases we have  + U +U− = C [T1 ] + C [T2 ], Pint D 2 C

[T1 ] [T2 ] ¯ − − C − = −λ δ T T2 T1 = −λ (T2− − T1− +

[T2 ] [T1 ] − ). 2 2

These relationships give U+ +U− = C [T1 ] + C [T2 ], P D 2   λ λ C C [T2 ] − [T1 ] = −λ (T2− − T1− ). + + T2− 2 T1− 2 

int

To simplify again, let us assume that λ is small compared to C/T . It results

162

10 Phase Change and Collisions



U+ +U− P D 2



int

= C [T1 ] + C [T2 ],

C C − − − [T2 ] − − [T1 ] = −λ (T2 − T1 ), T2 T1 which gives  +  U +U− + λ T2− (T2− − T1− ) , Pint D 2   T− U+ +U− C [T2 ] = − 2 − Pint D − λ T1− (T2− − T1− ) . 2 T2 + T1

C [T1 ] =

T1− − T2 + T1−

(10.16)

These formulas show that the mechanical work  + Ti− U +U− int P D ≥ 0, 2 T2− + T1− tends to warm up the two points and that the conduction effect  Ti−  − − − λ T (T − T ) , ± j 2 1 T2− + T1− tends to equalize their temperatures.

10.3 Experimental Results As far as we know, there are almost no thermal measurements in collisions. Nevertheless some results are available. For instance, temperature of the back face of a metallic plate collided by a steel ball has been measured with an infrared camera, [188, 189]. A temperature increase up to 7oC in less than a hundredth of a second is reported in Fig. 10.3. At the engineering scale, temperature is actually discontinuous, even if at a finer time scale it is continuous. If we assume the plate and the steel ball have the same temperature before collision, the formulas (10.16) within the small perturbation assumption give  C [θ ] = Pint D

U+ +U− 2

,

where C is the heat capacity of the steel ball, see also the examples of the following example . By assuming the ball has the same temperature than the plate, this result may be used to have information on the mechanical constitutive law.

10.4 The Thermomechanical Theory Involving Phase Change

163

Fig. 10.3 A metallic ball with 3 mm diameter collides a metallic plate with 0.85 mm thickness. The kinetic energy of the ball is 0.27 J. Temperature of the plate back face versus time is discontinuous at the time scale of the measurements, [189]

10.4 Collision of Two Balls: The Thermomechanical Theory Involving Phase Change Let us come back to the problem of warm rain falling on frozen ground or to the problem of a hailstone falling on the ground, i.e., to the thermal consequences of collisions of materials which can change phase. The heat source due to the collision can produce a very fast phase change concomitant to the collision, for instance the hailstone may melt. The volume fraction state quantities, βi , are involve. Let us recall that their evolution is described by an equation of motion. We assume the collision is instantaneous, thus it is wise to assume the phase change is also instantaneous: the very fast evolution of the volume fraction is replaced by a discontinuity. The functions βi (t) are bounded variations functions of time t. The discontinuities [βi ] are to be described by non-smooth equation of motion accounting for the microscopic motions.

10.4.1 The Equations of Motion The equations of motion are derived from the principle of virtual work which introduces different contributions. The actual and virtual velocities, U and V , are bounded variation functions of time t. The actual and virtual volume fractions, β = (β1 , β2 ) and b = (b1 , b2 ), are also bounded variation functions of time t. We denote

164

10 Phase Change and Collisions

   S(β ,t1 ,t2 ) = tk ∈ ]t1 ,t2 [  β + (tk ) = β − (tk ) , the set of discontinuity times of volume fractions β between times t1 and t2 (t1 < t2 ). The velocity differential measure  dβ =



dβ dt

dt +

∑

tk ∈S(β ,t1 ,t2 )

[β (tk )] δ (t − tk ),

is the sum of a Lebesgue measure, the smooth part of the differential, whose density is {dβ /dt} and of a Dirac measure, the non-smooth part of the differential, whose density at point tk is discontinuity [β (tk )] = β + (tk ) − β −(tk ). The set of the times when there is a collision, i.e., the set of the times when either U or β is discontinuous), is S(U, β ,t1 ,t2 ) = S(U,t1 ,t2 ) ∪ S(β ,t1 ,t2 ). The virtual work of the acceleration forces has already been given at the beginning of the chapter Tacc (t1 ,t2 ,U,V, β , b) =

2

∑

 t2

i=1 t1 2

+∑

i=1

mi

dUi (τ )Vi (τ )d τ dt



∑

tk ∈S(U,β ,t1 ,t2 )∪S(V,b,t1 ,t2

 Vi+ (tk ) + Vi− (tk ) , mi [Ui (tk )] 2 )

where m1 , m2 are the masses of the two balls. The virtual work of the interior forces is defined by Tint (t1 ,t2 ,U,V, β , b) =−

 t2 t1

Rint (τ )D(V )(τ ) + Bi (τ )bi (τ )d τ 

V + (tk ) + V − (tk ) − P (tk )D ∑ 2 tk ∈S(U,β ,t1 ,t2 )∪S(V,b,t1 ,t2 )





int

2

−∑

i=1

∑

tk ∈S(U,β ,t1 ,t2 )∪S(V,b,t1 ,t2 )

B pi (tk ) [bi (tk )] ,

where b = (b1 , b2 ) is the virtual velocity of phase change, B = (B1 , B2 ) and B p = (B p1 , B p2 ) are the interior microscopic works and percussion works. For the sake of completeness, we assume exterior work, Ai (t), and percussion work, Aip , with respect to the microscopic motion. The virtual work of the exterior forces is

10.4 The Thermomechanical Theory Involving Phase Change

165

Text (t1 ,t2 ,U,V, β , b)  2  t2 Fi (τ )Vi (τ ) + Ai (τ )bi (τ )d τ =∑ t1

i=1

2

+∑

i=1



∑

tk ∈S(U,β ,t1 ,t2 )∪S(V,b,t1 ,t2

  Vi+ (tk ) + Vi− (tk ) ext Pi (tk ) + A pi(tk ) [bi (tk )] , 2 )

with forces Fi , percussion Piext , work Ai and percussion work A pi . The principle of virtual work, gives easily the equations of motion at any time t m1 [U1 (t)] = −Pint (t) + P1ext (t), m2 [U2 (t)] = Pint (t) + P2ext (t), −B p1(t) + A p1(t) = 0,

(10.17)

−B p2(t) + A p2(t) = 0,

(10.18)

and almost everywhere m1

dU1 dU2 = −Rint + F1 , m2 = Rint + F2 , dt dt −B1 + A1 = 0, −B2 + A2 = 0.

Remark 10.8. Let us note that in the virtual works, the virtual macroscopic velocity intervene with (V + + V − ) /2 whereas the virtual microscopic velocities intervene with [b(tk )], because velocity U(t) is a discontinuous function of time and velocity dβ /dt is the derivative with respect to time of a discontinuous function of time. All the virtual works depend on the velocities. Set S(β ,t1 ,t2 ) actually depends on velocity differential measure dβ .

10.4.2 The Laws of Thermodynamics for Each Ball In the sequel we focus on the collisions and do not describe the smooth evolutions. Thus the equations are written for a unique collision.

10.4.2.1 The First Law The first law or energy balance for a ball, say ball number 1, is [E1 ] + [K1 ] = Text1 (t,t,U,U, β , β ) + C1,

(10.19)

166

10 Phase Change and Collisions

where   V + (t) + V1− (t) Text1 (t,t,U,V, β , b) = −Pint (t) + P1ext (t) 1 + A p1(t) [bi (t)], 2 is the virtual work of the exterior percussions applied to ball 1 and,     + − C1 = T1+ B1+ + B12 + T1− B1− + B12 = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 )) , is the heat impulse provided to the ball at the time of collision. This quantity includes the heat received from outside the system, T1 B1 , and the heat received from the inside, that is from the other ball, T1 B12 . Remark 10.9. The bold letter B represents an interior percussion work of damage whereas the calligraphic letter B represents an entropy impulse (une bouff´ee de chaleur in French). The theorem of expended energy for ball 1, [K1 ] = Tacc1 (t,t,U,U, β , β ) = Tint1 (t,t,U,U, β , β ) + Text1 (t,t,U,U, β , β ), where

Tint1 (t,t,U,U, β , b) = −B p1(t) [bi (t)].

With first law (10.19) we get [E1 ] = C1 − Tint1 (t,t,U,U, β , β ) = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 )) + B p1 [β1 ], (10.20) with

B+ − B− T+ +T− , Σ (B) = B + + B − , Δ (B) = . T¯ = 2 2

10.4.2.2 The Second Law Let be S1 the ball number 1 entropy, and recall the second law is + − + B12 = Σ (B1 ) + Σ (B12 ). [S1 ] ≥ B1+ + B1− + B12

By means of (10.20), the second law of thermodynamics gives [E1 ] − T¯1 [S1 ] ≤ [T1 ] (Δ (B1 ) + Δ (B12 )) + B p1 [β1 ], or, introducing the free energy Ψ = E − T S ,

(10.21)

10.4 The Thermomechanical Theory Involving Phase Change

[Ψ1 ] + S 1 [T1 ] ≤ [T1 ] (Δ (B1 ) + Δ (B12 )) + B p1 [β1 ].

167

(10.22)

Free energy of ball 1 is

Ψ1 (T1 , β1 ) = −C1 T1 ln T1 − β1

L1 (T1 − T0 ) + I(β1 ), T0

where C1 is the heat capacity, L1 is the latent heat at the phase change temperature T0 and the indicator function I of interval [0, 1] takes into account the internal constraint on the volume fraction 0 ≤ β1 ≤ 1. We define

L1 Ψˆ1 (T1 , β1 ) = −β1 (T1 − T0 ) + I(β1 ). T0

Then   [Ψ1 ] = [−C1 T1 ln T 1 ] + Ψˆ1 ,

(10.23)

where   + ˆ − − Ψˆ1 = Ψˆ1 (T + 1 , β 1 ) − Ψ1(T 1 , β 1 ) + ˆ + − ˆ + − ˆ − − = Ψˆ1 (T + 1 , β 1 ) − Ψ1(T1 , β1 ) + Ψ1 (T1 , β1 ) − Ψ1(T 1 , β 1 ),

and L1 Ψˆ1 (T1+ , β1− ) − Ψˆ1(T1− , β1− ) = −β1− [T1 ]. T0

(10.24)

In view of (10.23) and (10.24), inequality (10.22) transforms into [Ψ1 ] + S 1 [T1 ] L1 = [−C1 T1 ln T1 ] + Ψˆ1 (T1+ , β1+ ) − Ψˆ1(T1+ , β1− ) − β1− [T1 ] + S 1 [T1 ] T0 ≤ [T1 ] (Δ (B1 ) + Δ (B12 )) + B p1 [β1 ]. Since

(10.25)

[−C1 T1 ln T1 ] , [T1 ]

has a limit when [T1 ] −→ 0, we note  − L1 [T1 ] = −S1 [T1 ]. [−C1 T1 ln T1 ] + S 1 − β1 T0

(10.26)

168

10 Phase Change and Collisions

From (10.25), we have

Ψˆ1 (T1+ , β1+ ) − Ψˆ1 (T1+ , β1− ) ≤ [T1 ] (Δ (B1 ) + Δ (B12 ) + S1 ) + B p1 [β1 ]. (10.27) It is reasonable to assume that there is no dissipation with respect to [T1 ]

Δ (B1 ) + Δ (B12 ) + S1 = 0,

(10.28)

then, inequality (10.27) becomes

Ψˆ1 (T1+ , β1+ ) − Ψˆ1 (T1+ , β1− ) ≤ B p1 [β1 ].

(10.29)

10.4.3 The Laws of Thermodynamics for the System 10.4.3.1 The First Law The energy and entropy of the system are E = E1 + E2 + E int , S = S1 + S2 + S int , where E int and S int are the interaction energy and entropy. The first law for the system is [E ] + [K ] = Text (t,t,U,U, β , β ) + C , where C is the heat impulse provided to the system at the time of collision C = T1+ B1+ + T1− B1− + T2+ B2+ + T2− B2− . The theorem of expended energy, [K ] = Tacc (t,t,U,U, β , β ) = Tint (t,t,U,U, β , β ) + Text (t,t,U,U, β , β ), and the first law give [E ] = C − Tint (t,t,U,U, β , β ). The first law for every ball (10.20) and for the system (10.30) yield

(10.30)

10.4 The Thermomechanical Theory Involving Phase Change

169

  [E ] = [E1 ] + [E2 ] + E int = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 )) + B p1 [β1 ]   +T¯2 (Σ (B2 ) + Σ (B21 )) + [T2 ] (Δ (B2 ) + Δ (B21 )) + B p2 [β2 ] + E int = T1+ B1+ + T1− B1− + T2+ B2+ + T2− B2− 

U+ +U− +P D 2

+ B p1 [β1 ] + B p2 [β2 ]

int

= T¯1 Σ (B1 ) + [T1 ] Δ (B1 ) + T¯2 Σ (B2 ) + [T2 ] Δ (B2 ) 

U+ +U− +P D 2



int

+ B p1 [β1 ] + B p2 [β2 ].

It follows that 

E

int





U+ +U− =P D 2



int

−T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) − T¯2Σ (B21 ) − [T2 ] Δ (B21 ).

(10.31)

10.4.3.2 The Second Law The second law of thermodynamics is   [S ] = [S1 ] + [S2 ] + S int ≥ B1+ + B1− + B2+ + B2− = Σ (B1 ) + Σ (B2 ).

(10.32)

10.4.3.3 A Useful Inequality Let us note that if relationships 

 S int ≥ −Σ (B12 ) − Σ (B21 ) ,

(10.33)

and (10.21) are satisfied, the second law (10.32) is satisfied. Let us get an inequality equivalent to (10.33). To get this new inequality, we assume (10.33) is satisfied and transform relationship (10.31). Let be T¯1 + T¯2 ; Θ¯ = 2

170

10 Phase Change and Collisions

from (10.31) and (10.33), we have  +  int    U +U− − T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) E − Θ¯ S int ≤ Pint D 2 −T¯2 Σ (B21 ) − [T2 ] Δ (B21 ) + Θ¯ (Σ (B12 ) + Σ (B21 )) .

(10.34)

Introducing the free energy of interaction Ψ int = E int − Θ S int , inequality (10.34) can be written as  +  int  U +U− int int − T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) Ψ + S [Θ ] ≤ P D 2 −T¯2 Σ (B21 ) − [T2 ] Δ (B21 ) + Θ¯ (Σ (B12 ) + Σ (B21 )) , where S int

(10.35)

 int +  int − S + S [T1 ] + [T2 ] , [Θ ] = . = 2 2

Assuming Ψ int to be a smooth function of Θ , we have  int  Ψ + S int = 0. lim [Θ ]→0 [Θ ] Therefore, we can put  int  Ψ + S int [Θ ] = −Sint [Θ ],

(10.36)

and, from (10.35), we find  Sint U+ +U− ¯ − T1 Σ (B12 ) − [T1 ] Δ (B12 ) − 0≤P D 2 2  Sint −T¯2 Σ (B21 ) − [T2 ] Δ (B21 ) − + Θ¯ (Σ (B12 ) + Σ (B21 )). (10.37) 2 

int

It is reasonable to assume that there is no dissipation with respect to [T1 ] and [T2 ]

Δ (B12 ) −

Sint Sint = 0, Δ (B21 ) − =0; 2 2

(10.38)

then 

U+ +U− 0≤P D 2 int

− T¯1 Σ (B12 )

−T¯2 Σ (B21 ) + Θ¯ (Σ (B12 ) + Σ (B21 )).

(10.39)

10.4 The Thermomechanical Theory Involving Phase Change

171

But T¯1 Σ (B12 ) + T¯2 Σ (B21 )

Σ (B21 ) − Σ (B12 ) = Θ¯ (Σ (B12 ) + Σ (B21 )) + δ T¯ , 2

(10.40)

where the difference of temperature, δ T¯ , is defined by

δ T¯ = T¯2 − T¯1. Finally, inequality (10.39) yields  0 ≤ Pint D

U+ +U− 2

− δ T¯

Σ (B21 ) − Σ (B12 ) . 2

(10.41)

This last relationship links mechanical and thermal dissipations. We have proved that inequality (10.33) implies inequality (10.41). Let us prove the converse. Theorem 10.2. Let Ψ int be a smooth function of Θ . We assume no dissipation with respect to [T1 ] and [T2 ], (relationship (10.38)). If relationship (10.31) is satisfied, then (10.33) and (10.41) are equivalent. Proof. We assume (10.41) is satisfied. By means of (10.40) and supposing no dissipation with respect to [T1 ] and [T2 ], it follows that (10.37) is satisfied; thanks to (10.36), inequality (10.37) yields  +  int  U +U− int Ψ ≤P D − T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) 2 −T¯2 Σ (B21 ) − [T2 ] Δ (B21 ) + Θ¯ (Σ (B12 ) + Σ (B21 )) − S int [Θ ] . (10.42) Substituting the following identity  int   int    Ψ = E − [Θ ] S int − Θ¯ S int , into (10.42), we have 

−Θ S

int





≤− E

int





U+ +U− +P D 2



int

−T¯1 Σ (B12 ) − [T1 ] Δ (B12 ) − T¯2 Σ (B21 ) − [T2 ] Δ (B21 ) +Θ (Σ (B12 ) + Σ (B21 )) ;

(10.43)

172

10 Phase Change and Collisions

finally, by virtue of (10.31), (10.43) entails   −Θ S int ≤ Θ (Σ (B12 ) + Σ (B21 )) ,



that is, inequality (10.33) because temperatures are positive.

10.4.4 The Constitutive Laws For ball 1, we split the interior force B p1 into a possibly dissipative part due to the free energy, indexed by f e , and a dissipative part, indexed by d fe B p1 = B p1 + Bdp1.

The interior force, B1f e , is defined by free energy Ψ1 (T1 , β1 ), as follows fe B p1 ∈ ∂Ψ1 (T1+ , β1+ ) = −

L1 + (T − T0 ) + ∂ I(β1+), T0 1

(10.44)

where subdifferential ∂Ψ1 , is computed with respect to β1+ . Dissipative interior force, Bdp1 , is defined by Bdp1 ∈ ∂ Φ1 ([β1 ]) ,

(10.45)

Bdp1 [β1 ] ≥ 0,

(10.46)

and satisfies

where Φ1 ([β1 ]) is a pseudopotential of dissipation and the subdifferential is computed with respect to [β1 ]. For the system, we choose the following pseudopotential of dissipation   +   +   U +U− U +U− ¯ ¯ ,δT, χ = Φ D , δ T + I+(D U + ), Φ D 2 2 which depends on the velocity D ((U + + U − )/2), on the thermal heterogeneity δ T¯ and on the quantity χ = D (U − /2), which ensures the non-interpenetration condition, for instance, D (U + ) = U1+ − U2+ ≥ 0, in case ball 1 is on the right of ball 2. The constitutive laws are relationships (10.44), (10.45) and 

Σ (B21 ) − Σ (B12 ) P , 2 int



  + U +U− ¯ ∈ ∂Φ D ,δT,χ , 2

(10.47)

10.4 The Thermomechanical Theory Involving Phase Change

173

where the subdifferential set of Φ is computed with respect to the first two variables. Remark 10.10. From (10.47), applying the classical properties of pseudopotentials, we see that inequality (10.41) is satisfied. Moreover, the following theorems prove that the constitutive laws are such that the internal constraints and the second law are satisfied. Theorem 10.3. If the constitutive laws (10.44) and (10.45) are satisfied, then inequality (10.29) holds and the internal constraint 0 ≤ β + ≤ 1 is verified. Proof. If fe ∈ ∂Ψ1 (T1+ , β1+ ), B p1

then, using the definition of subdifferential, we have fe fe Ψˆ1 (T1+ , β1+ ) − Ψˆ1(T1+ , β1− ) ≤ B p1 (β1+ − β1−) = B p1 [ β1 ] ;

with (10.46) resulting from (10.45), the previous inequality entails fe Ψˆ1 (T1+ , β1+ ) − Ψˆ1 (T1+ , β1− ) ≤ B p1 [β1 ] + Bdp1 [β1 ] = B p1 [β1 ].

Since

∂Ψ1 (T1+ , β1+ ) = 0, / the internal constraint 0 ≤ β1+ ≤ 1 is verified.



Theorem 10.4. Assume no dissipation with respect to [T1 ] and [T2 ] (relationships (10.28) and (10.38)). If the first laws (10.20), (10.30) and the constitutive laws (10.44), (10.45) and (10.47) are verified, then the second law of thermodynamics holds for every ball and for the system. Proof. Due to Theorem 10.3, (10.23), (10.45) and (10.24), we have L1 [Ψ1 ] = [−C1 T1 ln T1 ] + Ψˆ1 (T1+ , β1+ ) − Ψˆ1 (T1+ , β1− ) − β1− [T1 ] T0 ≤ [−C1 T1 ln T1 ] + B p1 [β1 ] − β1−

L1 [T1 ]. T0

(10.48)

We have [Ψ1 ] = [E1 ] − [T1 S1 ] = [E1 ] − [T1 ] S 1 − T¯1 [S1 ], from (10.20), (10.48) and (10.49), it follows that

(10.49)

174

10 Phase Change and Collisions

L1 T¯1 [S1 ] ≥ [E1 ] − [T1 ] S 1 + [C1 T1 ln T1 ] − B p1 [β1 ] + β1− [T1 ] T0 = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 )) + B p1 [β1 ] − [T1 ] S 1 + [C1 T1 ln T1 ] − B p1 [β1 ] + β1−

L1 [T1 ] T0

= T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] (Δ (B1 ) + Δ (B12 ) + S1) , with S1 defined as in (10.26). From the previous inequality, assuming no dissipation with respect to [T1 ] (relationship (10.28)), we obtain [S1 ] ≥ Σ (B1 ) + Σ (B12 ) ,

(10.50)

that is, the second law of thermodynamics for ball 1. Due to constitutive laws (10.47) and Theorem 10.2, inequality (10.33) is satisfied. It results from (10.21) for each ball and from (10.33) that   [S ] = [S1 ] + [S2 ] + S int ≥ Σ (B1 ) + Σ (B12 ) + Σ (B2 ) + Σ (B21 ) − Σ (B12 ) − Σ (B21 ) = Σ (B1 ) + Σ (B2 ) ,



which is the second law for the system. Remark 10.11. The reaction to the internal constraint 0 ≤ non workless because ∂ I(β + )(β + − β −),

β+

≤ 1 is dissipative or

may be different from zero, in fact positive. Thus, the reaction is dissipative, as we may expect. This property is not true in a smooth evolution where the internal constraint is workless or non-dissipative because

∂ I(β )

dβ = 0, a.e.. dt

This is an example of the properties of the power of the reaction forces D reac given in Sect. 3.4.1.

10.5 Examples of Thermal Effects with Phase Changes Most of the following examples have been investigated by Anna Maria Caucci, [61]. We choose a pseudopotential of dissipation Φ , without non-diagonal terms, for example

10.5 Examples of Thermal Effects with Phase Changes

175

  +  − U +U− U ¯ ,δT,D Φ D 2 2   +  − λ U +U− U 2 = Φ mech D ,D + (δ T¯ ) . 2 2 4 With this choice, the mechanical problem is split up with the thermal one. The thermal constitutive law is

Σ (B21 ) − Σ (B12 ) = −λ δ T¯ .

(10.51)

Let us recall that the free energies chosen for the balls are

Ψi (Ti , βi ) = −Ci Ti ln T i − βi

Li (Ti − T0) + I(βi ), T0

i = 1, 2.

We choose the free energy of interaction Ψ int = 0 and have Sint = 0. and, from (10.38)

Δ (B12 ) = 0, Δ (B21 ) = 0.

It results from (10.28) that

Δ (B1 ) + S1 = 0, Δ (B2 ) + S2 = 0. Now, we suppose that the macroscopic mechanical problem is solved, that is, we know the quantity Pint D ((U + + U −) /2) and we assume no external percussion work A1 = A2 = 0. Due to equations of motion (10.17), (10.18), the thermal equations are [S1 ] = [C1 ln T1 ] +

T + L1 L1 [β1 ] = C1 ln 1− + [β1 ] = Σ (B1 ) + Σ (B12 ) , T0 T0 T1

[S2 ] = [C2 ln T 2 ] +

T + L2 L2 [β2 ] = C2 ln 2− + [β2 ] = Σ (B2 ) + Σ (B21 ) , T0 T0 T2

[E1 ] = C1 [T1 ] + L1 [β1 ] = T¯1 (Σ (B1 ) + Σ (B12 )) + [T1 ] Δ (B1 ) , [E2 ] = C2 [T2 ] + L2 [β2 ] = T¯2 (Σ (B2 ) + Σ (B21 )) + [T2 ] Δ (B2 ) , 

E

int





U+ +U− =P D 2 int

− T¯1 Σ (B12 ) − T¯2 Σ (B21 ) = 0.

176

10 Phase Change and Collisions

These equations are completed by the description of thermal relationships between the system and the outside and by the equations of microscopic motion fe + Bdpi ∈ − 0 = B pi = B pi

Li + (T − T0) + ∂ I(βi+ ) + ∂ Φi ([βi ]). T0 i

10.5.1 The Thermal Equations in the Isentropic Case If the collision is isentropic

Σ (B1 ) = Σ (B2 ) = 0, and the thermal equations become

Δ (B12 ) = 0, Δ (B21 ) = 0, Δ (B1 ) + S1 = 0, Δ (B2 ) + S2 = 0, T+ C1 ln 1− T1 C2 ln

+

(10.52)

L1 [β1 ] = Σ (B12 ) , T0

T2+ L2 + [β2 ] = Σ (B21 ) , T2− T0

C1 [T1 ] + L [β1 ] = T¯1 Σ (B12 ) + [T1 ] Δ (B1 ) ,

(10.53)

C2 [T2 ] + L [β2 ] = T¯2 Σ (B21 ) + [T2 ] Δ (B2 ) ,

(10.54)

 Pint D

U+ +U− 2

= T¯1 Σ (B12 ) + T¯2 Σ (B21 ) .

(10.55)

Substituting the expressions of Σ (B12 ) and Σ (B21 ) in (10.51) and (10.55), we obtain C2 ln

+ T2+ L1 ¯2 = C1 ln T1 + L2 [β1 ] + λ δ T¯1 , + [ β ] + λ T 2 T2− T0 T1− T0

 Pint D

U+ +U− 2



 T + L1 = T¯1 C1 ln 1− + [β1 ] T0 T1  T + L2 +T¯2 C2 ln 2− + [β2 ] . T0 T2

(10.56)

(10.57)

10.5 Examples of Thermal Effects with Phase Changes

177

Equations (10.53) and (10.54) give the heat exchanges, Δ (B1 ) and Δ (B2 ), with the outside, while (10.56) and (10.57) give the temperatures T1+ and T2+ after the collision. The splitting of the heat exchanges with the exterior, the B + and B− is given by (10.52).

10.5.2 The Thermal Equations in the Adiabatic Case If the collision is adiabatic T1+ B1+ + T1− B1− = T¯1 Σ (B1 ) + [T1 ] Δ (B1 ) = 0, T2+ B2+ + T2− B2− = T¯2 Σ (B2 ) + [T2 ] Δ (B2 ) = 0, and we have the following equations

Δ (B12 ) = 0, Δ (B21 ) = 0, Δ (B1 ) + S1 = 0, Δ (B2 ) + S2 = 0, [S1 ] = C1 ln

T1+ L1 + [β1 ] = Σ (B1 ) + Σ (B12 ) , T1− T0

[S2 ] = C2 ln

T2+ L2 + [β2 ] = Σ (B2 ) + Σ (B21 ) , T2− T0

C1 [T1 ] + L1 [β1 ] = T¯1 Σ (B12 ) , C2 [T2 ] + L2 [β2 ] = T¯2 Σ (B21 ) ,  + U +U− Pint D = C1 [T1 ] + L1 [β1 ] + C2 [T2 ] + L2 [β2 ], (10.58) 2 [T2 ] L2 [T1 ] L1 C2 ¯ + ¯ [β2 ] − C1 ¯ − ¯ [β1 ] = −λ δ T¯ . T2 T2 T1 T1

(10.59)

The last two equations give the temperatures T1+ and T2+ after the collision, the first give the entropy exchanges, Σ (B1 ) and Σ (B2 ), with the outside.

10.5.3 The Adiabatic Situation with the Small Perturbation Assumption We consider the adiabatic case and we assume small perturbations, that is   Ti± = T0 + θi± , θi±  T0 .

178

10 Phase Change and Collisions

We can write (10.58) and (10.59) as follows  Pint D

C2

U+ +U− 2

= C1 [θ1 ] + L1 [β1 ] + C2 [θ2 ] + L2 [β2 ],

(10.60)

 [θ2 ] L1 [θ1 ] L2 [θ2 ] [θ1 ] − − , − + [β2 ] − C1 − [ β 1 ] = − λ θ2 − θ1 + T0 T0 T0 T0 2 2 (10.61)

and get the system U+ +U− − 2L1 [β1 ] 2  [θ2 ] [θ1 ] − − , − + λ T0 θ2 − θ1 + 2 2  + U +U− − 2L2 [β2 ] 2C2 [θ2 ] = Pint D 2  [θ2 ] [θ1 ] − − − λ T0 θ2 − θ1 + , − 2 2 

2C1 [θ1 ] = Pint D

(10.62)

(10.63)

where the volume fractions β1 and β2 satisfy the equations of microscopic motion 0∈−

L1 + (T − T0) + ∂ I(β1+ ) + ∂ Φ1 ([β1 ]) , T0 1

(10.64)

0∈−

L2 + (T − T0) + ∂ I(β2+ ) + ∂ Φ2 ([β2 ]) , T0 2

(10.65)

which are equivalent to

θ1+ ∈ ∂ I(β1+ ) +

T0 ∂ Φ1 ([β1 ]) , L1

(10.66)

θ2+ ∈ ∂ I(β2+ ) +

T0 ∂ Φ2 ([β2 ]). L2

(10.67)

It is easy to prove that (10.62), (10.66) and (10.63), (10.67) giving the temperatures and phase fractions θ1+ , θ2+ and β1+ , β2+ have unique solutions.

10.5 Examples of Thermal Effects with Phase Changes

179

10.5.4 Identical Balls at the Same Temperature Before the Collision The temperatures after the collision are equal. Equations (10.62) and (10.66) show that they do not depend on λ .

10.5.4.1 Case Where the Phase Change Is Not Dissipative We consider two identical balls of ice, C1 = C2 = C, L1 = L2 = L at the same temperature, θ − , before the collision. The balls have the same temperature and volume fraction after the collision, θ + , β +

θ1− = θ2− = θ − , θ1+ = θ2+ = θ + ,

(10.68)

β1−

(10.69)

=

β2−

= 0,

β1+

=

β2+

+

=β .

We assume no dissipation with respect to the volume fractions discontinuities [β1 ] and [β2 ]

Φ1 ([β1 ]) = Φ2 ([β2 ]) = 0.

(10.70)

Equations (10.62) and (10.63) give C [θ ] = where

 1 T − 2Lβ + , 2

(10.71)

U+ +U− . T =P D 2 

int

Because of (10.70), we find easily from the equations of microscopic motion (10.66), (10.67)

θ + ∈ ∂ I(β + ).

(10.72)

The evolution of the balls is described in the following theorem Theorem 10.5. We have 1 if T ≤ −2Cθ −, then β + = 0: after collision there is ice with temperature given by (10.71); 2 if T ≥ 2 (L − Cθ −), then β + = 1: after collision there is liquid water with temperature given by (10.71); 3 if −2Cθ − < T < 2 (L − Cθ − ), then 0 < β + < 1: after collision there is a mixture of ice and liquid water with temperature θ + = 0. Volume fraction β + is given by (10.71).

180

10 Phase Change and Collisions

Proof. 1 If T ≤ −2Cθ − , then, from (10.71), it follows that C [θ ] + Lβ + ≤ −Cθ − . The previous inequality entails L θ+ ≤ − β+ ≤ 0 ; C therefore, in view of (10.72), we obtain

β + = 0. 2 Let be T ≥ 2 (L − Cθ − ). We get C [θ ] + Lβ + ≥ L − Cθ − , that is

θ+ ≥

 L 1 − β + ≥ 0. C

On the other hand, because θ + ∈ ∂ I(β + ), we get

β + = 1. 3 If −2Cθ − < T < 2 (L − Cθ − ), then −

 L + L 1−β+ . β < θ+ < C C

(10.73)

Now, if θ + > 0, because of (10.72), we have β + = 1 and, substituting into (10.73)

θ+ < 0 ; analogously, if θ + < 0, we find β + = 0 and, from (10.73)

θ + > 0. Finally, if we suppose θ + = 0, then (10.73) is satisfied and constitutive law (10.72) implies 0 < β + < 1. Volume fraction is given by (10.71)

β+ =

T + 2Cθ − . 2L

10.5 Examples of Thermal Effects with Phase Changes

181

The result of the theorem agrees with what is expected: a violent collision produces a phase change whereas a non violent collision does not. Violent means dissipative, i.e., T large.

10.5.4.2 Case Where the Phase Change Is Dissipative For every ball, let us choose the pseudopotential of dissipation as c Φi ([βi ]) = [βi ]2 , i = 1, 2, 2 where c is a positive constant. Thus, equations of motion (10.64) and (10.65) give L + θ ∈ ∂ I(β1+ ) + c [β1 ], T0 1 L + θ ∈ ∂ I(β2+ ) + c [β2 ], T0 2 and we have the equations C [θ ] =

 1 T − 2Lβ + , 2

L + θ ∈ ∂ I(β + ) + cβ +. T0 The following result is easily proved Theorem 10.6. We have 1 if T ≤ −2Cθ −, then β + = 0: after collision there is solid with θ + ; 2 if T ≥ 2 (L − Cθ − + cCT0 /L), then β + = 1: after collision there is liquid with θ + ≥ (cT0 )/L; 3 if −2Cθ − < T < 2 (L − Cθ − + cCT0 /L), then 0 < β + < 1: after collision there is a mixture of solid and liquid with θ + = (cT0 β + )/L. When there is dissipation the collision has to be more violent to melt the solid balls. The phase change occurs with temperature slightly above T0 , as it is the case for dissipative phase changes, see Sect. 4.7.

10.5.5 Identical Balls at Different Temperatures Before the Collision: Collision of Two Pieces of Ice When two pieces of ice at different temperature collide, the dissipation due to the collision may be large enough to melt the warmest of them. We look for conditions

182

10 Phase Change and Collisions

on the state quantities before the collision and on the dissipated work, such that this phenomenon occurs. We expect that the temperatures before collision cannot be very cold and that the dissipated work has to be large. We assume that there is no dissipation with respect to [β1 ] and [β2 ]

Φ1 ([β1 ]) = Φ2 ([β2 ]) = 0, thus the equations of microscopic motion are (10.66), (10.67)

θ1+ ∈ ∂ I(β1+ ), θ2+ ∈ ∂ I(β1+ ).

(10.74)

The two identical pieces of ice before collision satisfy

β1− = β2− = 0, θ1− ≤ 0, θ2− ≤ 0. We look for conditions such that ball 1 melts and ball 2 remains frozen

β1+ = 1, β2+ = 0.

(10.75)

θ1+ ≥ 0, θ2+ ≤ 0.

(10.76)

Thus from (10.74), we have

10.5.5.1 Case with λ Small with Respect to C/T0 Equations (10.19) and (10.27) give 

U+ +U− P D 2 int



C2 λ + T0 2



= C1 [θ1 ] + L [β1 ] + C2 [θ2 ] + L [β2 ],

   L L C1 λ [θ2 ] + [β2 ] − [θ1 ] − [β1 ] = −λ θ2− − θ1− . + T0 T0 2 T0

We assume λ small with respect to Ci /T0 , i = 1, 2 (see Remark 10.12) and get  Pint D

U+ +U− 2

= C1 [θ1 ] + L [β1 ] + C2 [θ2 ] + L [β2 ],

  C2 L C1 L [θ2 ] + [β2 ] − [θ1 ] − [β1 ] = −λ θ2− − θ1− , T0 T0 T0 T0

10.5 Examples of Thermal Effects with Phase Changes

183

which gives the system C1 [θ1 ] =

  +   1 U +U− Pint D − 2L [β1 ] + λ T0 θ2− − θ1− , 2 2

(10.77)

C2 [θ2 ] =

  +   1 U +U− Pint D − 2L [β2 ] − λ T0 θ2− − θ1− , 2 2

(10.78)

where the volume fractions β1 and β2 satisfy the equations of microscopic motion (10.74). Equations (10.77), (10.78) give with C1 = C2 = C, and (10.75) C [ θ1 ] =

  1 T − 2L + λ T0 θ2− − θ1− , 2

(10.79)

C [ θ2 ] =

  1 T − λ T0 θ2− − θ1− . 2

(10.80)

By means of (10.79), (10.80), conditions (10.76) are satisfied if and only if (2C − λ T0 ) θ1− + λ T0 θ2− + T − 2L ≥ 0,

λ T0 θ1− + (2C − λ T0 ) θ2− + T ≤ 0,

(10.81)

with θ1− ≤ 0, θ2− ≤ 0. Because of our hypothesis on λ , we have 2C − λ T0 > 0.

(10.82)

If T < 2L, inequality (10.81)1 and (10.82) show that it is impossible to satisfy system (10.81) with θ1− ≤ 0, θ2− ≤ 0. Thus if the dissipation is small, it is impossible to melt one piece of ice. Both of them remain frozen. If T ≥ 2L, it is possible to find (θ1− , θ2− ) satisfying system (10.81) with θ1− ≤ 0, − θ1 0 (θ1− = 0 if T = 2L) and θ2− < θ1− . Thus if the dissipation is large, one piece of ice melts, the other one remains frozen. The temperature of the coldest piece of ice has to be sufficiently cold. Examples are given on Figs. 10.4 and 10.5 for λ = 0 J / K2 and λ = 1000 J / K2 , T = 10 L, with L = 3.33 × 105 J, C = 106 J / K and T0 = 273 K. 10.5.5.2 Case When λ Is Not Small with Respect to C/T0 We have [β1 ] = 1, [β2 ] = 0,

184

10 Phase Change and Collisions

Fig. 10.4 The case λ = 0 J / K2 and T = 10 L, with L = 3.33 × 105 J, C = 106 J / K. The inequalities (10.81) have solutions if θ1− is negative and satisfies θ1− ≥ −(T − 2L)/2C, equality on the blue line and θ2− ≤ −T /2C, equality on the green line. The point (θ1− , θ2− ) has to belong to set M

Equations (10.60), (10.61) give T = C [θ1 ] + L + C [θ2 ] ,  [θ1 ] L [θ2 ] [θ1 ] [θ2 ] − − . − −C − = −λ θ2 − θ1 + C T0 T0 T0 2 2

(10.83) (10.84)

From (10.83), we get [ θ2 ] =

1 (T − C [θ1 ] − L) ; C

(10.85)

substituting (10.85) into (10.84), we obtain

θ1+ =

λ T0 L 2C T −L − θ1− + θ2− + , 2C + λ T0 2C + λ T0 2C 2C + λ T0

θ2+ =

L 2C T −L λ T0 + θ− + θ− + . 2C + λ T0 2 2C + λ T0 1 2C 2C + λ T0

(10.86)

10.5 Examples of Thermal Effects with Phase Changes

185

Fig. 10.5 The case λ = 1000 J / K2 and T = 10 L, with L = 3.33 × 105 J, C = 106 J / K. The inequalities (10.81) have solutions if point (θ1− , θ2− ) belongs to set M which is defined by the blue and green lines of (10.81) and θ1− ≤ 0, θ2− ≤ 0

Conditions (10.76) are satisfied if and only if T −L λ T0 L 2C θ− + θ− + ≥ 0, − 2C + λ T0 1 2C + λ T0 2 2C 2C + λ T0

λ T0 2C L T −L + θ2− + θ1− + ≤ 0. 2C + λ T0 2C + λ T0 2C 2C + λ T0

(10.87)

Because θ1− ≤ 0, θ2− ≤ 0, to satisfy (10.87)1 a necessary condition is L T −L ≥ . 2C 2C + λ T0 Thus the dissipation has to be large in order to melt one of the pieces of ice. If λ < 2C/T0 , then system (10.87) has solutions (θ1− , θ2− ) such that θ1− ≤ 0, θ2− ≤ 0, if T −L L L ≥ > , 2C 2C − λ T0 2C + λ T0 see Figs. 10.6 and 10.7. This relationship is valid when λ is negligible. If λ is small, only the mechanical effect warms the balls whereas the conduction has a negligible effect. We have

186

10 Phase Change and Collisions

Fig. 10.6 For λ < C/(2T0 ), the first condition of (10.87), curve (1) and the second condition of (10.87), curve (2). In order to have the ball number 1 unfrozen after collision while the ball number 2 remains frozen, the temperatures θ1− and θ2− have to be in the hatchered triangle

Fig. 10.7 The curve y(λ ) = L/(2C − λ T0 ) versus λ in blue and the curve y = L/(2C + λ T0 ) in green for λ < 2C/T0 . The quantity (T − L) /2C has to be in the blue domain M for the ball number 1 to melt. Thus the collision has to be violent enough, i.e., T large enough for the ball number 1 to melt

10.5 Examples of Thermal Effects with Phase Changes

187

Fig. 10.8 For λ > C/(2T0 ), the first condition (10.87), curve (1) and the second condition of (10.87), curve (2). In order to have the ball number 1 unfrozen after collision while the ball number 2 remains frozen, the temperatures θ1− and θ2− have to be in the hatchered triangles

θ2− ≤ θ1− ≤ 0, and θ2+ ≤ 0 ≤ θ1+ . The warmest piece of ice melts in the collision and the coldest remains frozen. If λ > 2C/T0 , then system (10.87) has solutions such that θ1− ≤ 0, θ2− ≤ 0 if L L T −L > , ≥− 2C 2C − λ T0 2C + λ T0 see Figs. 10.8 and 10.9. If C is small, i.e., if the heat capacity is negligible, it is difficult for the system to store energy (the only possibility to store energy is with change phase). Because we have assumed the system to be adiabatic, the heat has to remain in the system and very large temperature variations occur: the effect of conduction is added to the mechanical effect and it increases the temperature of the coldest ball. Therefore, the coldest ball before collision becomes the warmest and vice versa. We have

θ1− ≤ θ2− ≤ 0, and θ2+ ≤ 0 ≤ θ1+ . In the extreme situation where λ = ∞, formulas (10.86) show that

θ1+ = θ2− , θ2+ = θ1− . When the thermal dissipation is very large the temperatures exchange. Let us note the same result holds for the velocities with a quadratic mechanical pseudopotential of dissipation (10.4). When the mechanical dissipation parameter k = ∞, formulas (10.8) show that the velocities exchange U1+ = U2− , U2+ = U1− , in case m1 = m2 .

188

10 Phase Change and Collisions

Fig. 10.9 The curve y(λ ) = −L/(2C − λ T0 ) versus λ in red and the curve y = L/(2C + λ T0 ) in green for λ > 2C/T0 . The quantity (T − L) /2C has to be in the blue domain M for the ball number 1 to melt. Thus the collision has to be violent enough, i.e., T large enough for the ball number 1 to melt

In the extreme situation where the heat capacity is zero, C = 0, the system cannot store energy except by changing phase. The energy L which is needed to melt the piece of ice has to be equal to the dissipated work T and the discontinuities of temperature are opposite due to (10.83) extended by continuity. Remark 10.12. The assumption λ > 2C/T0 is not often realistic because λ is proportional to the contact surface of the two pieces of ice and C is proportional to the volume of the pieces of ice. To have λ > 2C/T0 , the contact surface has to be large compared to the volume. For instance the two pieces of ice are planes with small thickness. But in this case the contact surface with the atmosphere is also large in contradiction with the adiabatic assumption in collision. Thus the assumption λ > 2C/T0 is not coherent when assuming adiabatic collisions, i.e., collisions without heat exchange with the exterior. If λ = 2C/T0 , then, from (10.84), we have 

this entails

C λ + T0 2



    + C λ  − L L θ2 − θ1+ = + − θ2 − θ1− = ; T0 T0 2 T0

θ2+ > θ1+ ,

10.5 Examples of Thermal Effects with Phase Changes

189

which forbids to have θ1+ ≥ 0 ≥ θ2+ . In this situation, the effect of conduction and heat storage have opposite effects and cancel each other.

10.5.6 Identical Balls at Different Temperatures Before the Collision: Collision of a Droplet of Water with a Piece of Ice We assume that λ is small with respect to C/T0 and that there is no dissipation with respect to [β1 ] and [β2 ]

Φ1 ([β1 ]) = Φ2 ([β2 ]) = 0. A droplet of rain and a piece of ice collide. Before collision we have

β1− = 1, β2− = 0, θ1− ≥ 0, θ2− ≤ 0. We look for conditions such that the droplet of rain freezes and the ice remains frozen

β1+ = 0, β2+ = 0.

(10.88)

θ1+ ≤ 0, θ2+ ≤ 0.

(10.89)

Thus from (10.74), we have

We assume that the collision is adiabatic and to be coherent that λ is small with respect to C/T0 . Equations (10.62), (10.63) give with C1 = C2 = C, using and (10.88) C [ θ1 ] =

  1 T + 2L + λ T0 θ2− − θ1− , 2

C [ θ2 ] =

  1 T − λ T0 θ2− − θ1− . 2

By means of (10.79), (10.80), conditions (10.89) are satisfied if and only if (2C − λ T0 ) θ1− + λ T0 θ2− + T + 2L ≤ 0,

λ T0 θ1− + (2C − λ T0 ) θ2− + T ≤ 0,

(10.90)

with θ1− ≥ 0, θ2− ≤ 0. Because of the hypothesis 2C − λ T0 > 0, it is always possible to satisfy conditions (10.90) by having θ2− negative enough, i.e., by having the ice piece cold enough. The maximum value of θ2− is given by θ1+ = 0. It is

190

10 Phase Change and Collisions

θ2− = −

T + 2L 2C − λ T0 − , λ T0 λ T0

or

θ2− = −

T + 2L 2C − − θ , λ T0 λ T0 1

because λ is small with respect to C/T0 . As it may be expected, the maximum value of θ2− is negative and decreasing when the dissipated T work is increasing and when θ1− is increasing. These results have been reported in [62].

10.5.7 A Droplet of Rain Falls on a Frozen Ground: Does It Give Black Ice? Let us consider a droplet of rain falling on a frozen ground and keep the assumption that the two solid colliding have uniform temperatures. Because the heat capacities of the media are proportional to their volumes, it is wise to have C2 = ∞ (the soil is solid number 2). It results from (10.63) that indeed the temperature of the ground is not modified by the collision of the water droplet. It remains frozen

θ2+ = θ2− , β2+ = β2− . The two remaining equations are C1 [θ1 ] =

  1 [θ1 ] T − 2L1 [β1 ] + λ T0 θ2− − θ1− − , 2 2

θ1+ ∈ ∂ I(β1+ ), with

(10.91)

θ1− > 0, β1− = 1, θ2− < 0.

They give L1 β1+ + ∂ I(β1+ ) 

  1 λ T0 − T + 2L1 + λ T0 θ2− − θ1− + (C1 + )θ1 = A . 2 4

The solution of the system (10.91) is: • If A ≥ L1 , β1+ = 1. The droplet remains liquid and there is no black ice. This situation requires that   1 λ T0 − T + λ T0 θ2− − θ1− + (C1 + )θ1 ≥ 0, 2 4

10.5 Examples of Thermal Effects with Phase Changes

or

191

λ T0 − λ T0 − T θ + (C1 − + )θ1 ≥ 0. 2 2 2 4

Assuming C1 − λ T0 /4 ≥ 0, the only quantity which is negative is (λ T0 /4)θ2− . Thus the temperature of the frozen ground has to be not too cold to avoid black ice. In the same way, warm droplets and very dissipative collisions avoid black ice. After colliding the soil, the liquid water with temperature

θ1+ = θ1− +

1 2C1 +

λ T0 2

   T + λ T0 θ2− − θ1− ≥ 0,

is in contact with a cold layer. The equations of Chap. 9 are to be used to describe its evolution. • If A ≤ 0, β1+ = 0. The droplet freezes and there is black ice. This situation requires that   1 λ T0 − T + 2L1 + λ T0 θ2− − θ1− + (C1 + )θ1 ≤ 0, 2 4 or

T λ T0 − λ T0 − + L1 + )θ1 ≤ 0. θ + (C1 − 2 2 2 4 The temperature of the ground has to be cold enough to get black ice with temperature

θ1+ = θ1− +

1 2C1 +

λ T0 2

   T + 2L1 + λ T0 θ2− − θ1− ≤ 0.

• If 0 < A < L1 , 0 < β1+ < 1. The droplet freezes partially and its temperature is θ1+ = 0. The unfrozen water volume fraction is

β1+ =

A . L1

After colliding the soil, the evolution of the mixture of ice and water is described by the equations of Chap. 9.

•

Chapter 11

Collisions of Deformable Bodies and Phase Change

Let us consider again a piece of ice colliding with a warm soil. Work is mostly dissipated on the contact surface and in its neighbourhood. Thus the temperature discontinuity is more important in this zone and some melting may occur there and not elsewhere. In this chapter, collisions of deformable solids with volume discontinuities of velocity are investigated together with the resulting phase changes. We introduce percussions related to volume velocity discontinuities and to phase volume fraction discontinuities. We assume the contact surface of the colliding solids is conformal, i.e., it has a nonzero Lebesgue measure.

11.1 The Principle of Virtual Work The virtual work of the interior percussion forces, is Tint (Ω1 ∪ Ω 2 , V, γ ) = − − − −

   

Ω1

Σ1 : D(

∂ Ω1 ∩∂ Ω 2 Ω1 Ω2

− V+ 1 + V1 )d Ω 1 − 2

R p · Ds (

 Ω2

Σ2 : D(

− V+ 2 + V2 )d Ω2 2

V+ + V− )d Γ 2

B p1 [γ1 ] + H p1 · grad[γ1 ] d Ω1 B p2 [γ2 ] + H p2 · grad[γ2 ] d Ω2 ,

where the virtual velocities are V = (V1 , V2 ), γ = (γ1 , γ2 ), the D(V1 ), D(V2 ) are the usual strain rates and Ds (V) = V2 − V1 is the gap velocity. The domains occupied by the two solids at time t are Ω1 (t) = Ω1 and Ω 2 (t) = Ω 2 . It appears, contact percussions R p , percussion stresses Σ and, percussion works B, percussion work flux vectors H p . The virtual work of the acceleration forces is M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 11, © Springer-Verlag Berlin Heidelberg 2012

193

194

11 Collisions of Deformable Bodies and Phase Change

Tacc (Ω1 ∪ Ω2 , V, γ ) =

 Ω1

ρ1 [U1 ] ·

− V+ 1 + V1 d Ω1 + 2

 Ω2

ρ2 [U2 ] ·

− V+ 2 + V2 d Ω2 . 2

Remark 11.1. The mass balance equations are [ρ ] = 0. When the material are incompressible, the incompressibility conditions are taken into account by the pseudopotentials of dissipation introducing percussion pressures, [100, 115]. The exterior forces have the virtual work Text (Ω1 ∪ Ω2 , V, γ ) =



F p1 ·

Ω1

+ + +

 ∂ Ω1



Ω1

− V+ 1 + V1 d Ω1 + 2

T p1 ·

− V+ 1 + V1

2

A 1 [ γ1 ] d Ω 1 +



∂ Ω1

a1 [γ1 ] d Γ1 +

 Ω2

d Γ1 +

 Ω2

F p2 ·

 ∂ Ω2

− V+ 2 + V2 d Ω2 2

T p2 ·

− V+ 2 + V2 d Γ2 2

A 2 [ γ2 ] d Ω 2



∂ Ω2

a2 [γ2 ] dΓ2 .

We assume that the surface exterior percussions T p are applied to the whole boundary of each solid (they can be equal to zero in some parts of the boundaries). The F p are the volume exterior percussions. The A and a are the volume and surface percussion work provided by the exterior by electrical, radiative,... actions. The equations of motion results from the principle of virtual work ∀V, ∀γ , Tacc (Ω1 ∪ Ω2 , V, γ ) = Tint (Ω1 ∪ Ω2 , V, γ ) + Text (Ω1 ∪ Ω2 , V, γ ). Different choices of the virtual velocities V = (V1 , V2 ) and γ = (γ1 , γ2 ) give

ρ1 [U1 ] = div Σ1 + F p1 , −B p1 + divH p1 + A1 = 0, in Ω1 ,

(11.1)

ρ2 [U2 ] = div Σ2 + F p2 , −B p2 + divH p2 + A2 = 0, in Ω2 ,

(11.2)

Σ1 N1 = R p + T p1 , H p1 · N1 = a1 , on ∂ Ω 1 ∩ ∂ Ω 2 , Σ2 N2 = −R p + T p2 , H p2 · N2 = a2 , on ∂ Ω1 ∩ ∂ Ω2 , Σ1 N1 = T p1 , H p1 · N1 = a1 , on ∂ Ω1 \ (∂ Ω1 ∩ ∂ Ω2 ) , Σ2 N2 = T p2 , H p2 · N2 = a2 , on ∂ Ω 2 \ (∂ Ω 1 ∩ ∂ Ω 2 ) .

11.2 The First Law of Thermodynamics

195

11.2 The First Law of Thermodynamics As in the previous chapter, we define [T2 ] = T2+ − T2− , [T1 ] = T1+ − T1− ; those two quantities are analogous to the velocity dT /dt in a smooth evolution, and T2 =

T2+ + T2− T + + T1− , T1 = 1 , 2 2

in Ω1 and Ω2 , T2 + T1 [T1 ] + [T2 ] , [Θ ] = Θ + − Θ − = , 2 2  1 + T + T1 Θ+ +Θ− T + T1+ + T2− + T1− , = = = 2 2 2 4 2

δ T = T 2 − T 1, Θ = Tm

in ∂ Ω1 ∩ ∂ Ω2 . Quantity δ T at point x1 = x2 = x ∈ ∂ Ω1 ∩ ∂ Ω2 is analogous to grad T in a smooth situation. The states quantities of the solids are E1 = (T1 , β1 , grad β1 , ε1 ) and E2 = (T2 , β2 , grad β2 , ε2 ), where the deformation ε are continuous in the collision. The state quantities of the contact surface are E ± = Θ on ∂ Ω 1 ∩ ∂ Ω 2 . The evolutions of the two solids are described by

Δ E1± = (D(

− U+ 1 + U1 ), [β1 ] , grad [β1 ] , grad T 1 ) in Ω1 , 2

Δ E2± = (D(

− U+ 2 + U2 ), [β2 ] , grad [β2 ] , grad T 2 ) in Ω2 . 2

The evolution of the contact surface is describe by

δ E ± = (Ds (

U+ + U− ), δ T ) on ∂ Ω1 ∩ ∂ Ω 2 . 2

The free energies of the two solids are Ψ1 (E1 ) and Ψ2 (E2 ). The structure of the expression of the work of the interior percussions leads us to introduce surface free energies, Ψ1s , Ψ2s . Now let us consider the thermal effects. We focus on the basic problem, the collision of two solids. Thus the equations which are considered are relative to a collision which occurs at time t. For the sake of simplicity, time t is deleted from the formulas. Following our rules, the heat intakes are equal to the entropy intakes multiplied by the temperatures at which the entropies are received, either at temperature T + or at temperature T − . Let D1 ∪ D2 be a subdomain of Ω1 ∪ Ω2 . Let

196

11 Collisions of Deformable Bodies and Phase Change

us denote O1 = ∂ Ω1 ∩ ∂ D1 , O2 = ∂ Ω2 ∩ D2 , and O = O1 ∩ O2 . When the solids collide, we assume that each solid receives heat impulses from the other through local actions. The heat impulse received by subdomain D1 ∪ D2 is Q (D1 ∪ D2 ) =

 D1

+ + =

Σ (T1 B1 )d Ω1 +

 ∂ D1 \O



∂ D2 \O



+ −

 ∂ D1



O

D2

Σ (T2 B2 )d Ω2

Σ (T1 Q p1 )dΓ1 + Σ (T2 Q p2 )dΓ2 +

Σ (T1 B1 )d Ω1 +

D1



 D2



Σ (T1 Bs1 )d Γ1

∂ D1



Σ (T2 Bs2 )d Γ2

∂ D2

Σ (T2 B2 )d Ω2 

Σ (T1 (Q p1 + Bs1 )) d Γ1 +

∂ D2

Σ (T2 (Q p2 + Bs2 )) dΓ2

Σ (T1 Q p1 + T2 Q p2 ) dΓ ,

where Σ (X) = X + + X −. The T Q p ’s are the heat impulses supplied by contact action from the exterior of the system to the surfaces, and the T Bs ’s are the exterior surface heat impulses, for instance, sources due to a chemical rapid reaction (the heat T Bs is present everywhere in the preceding expression, whereas the heat supplied by the exterior T Q p is absent from part ∂ D1 ∩ ∂ D2 where there is no heat supplied from the exterior by contact). The T B’s are the volume heat impulses received from the exterior. The functions T Q p are defined on ∂ Ω 1 ∩ ∂ Ω2 : they are the heat supplied by one solid to another one by contact (this property can be seen by choosing one of the subdomains D empty). The structure of the expression of the work of the interior forces and the structure of the expression of the received heat lead us to assume surface internal energies es besides the volume internal energies. The energy balance of the subdomain D1 ∪ D2 at any time t, is  D1 ∪D2

[e]d Ω +

 O1

[es1 ]dΓ1 +

 O2

[es2 ]d Γ2 = −Tint (D1 ∪ D2 , U, β ) + Q (D1 ∪ D2 ) ,

where Tint (D1 ∪ D2 , U, β ) is the actual work of the percussions interior to the subdomain D1 ∪ D2 . We get  D1 ∪D2

[e]d Ω +

 O1

[es1 ]dΓ1 +

 O2

[es2 ]dΓ2 =

 O

+

R p · Ds (

 D1

U+ + U− )d Γ 2

Σ1 : D(

− U+ 1 + U1 ) 2

+B p1 [β1 ] + H p1 · grad[β1 ] d Ω1

11.2 The First Law of Thermodynamics

197

+

 D2

Σ2 : D(

− U+ 2 + U2 ) 2

+B p2 [β2 ] + H p2 · grad[β2 ] d Ω2 +Q (D1 ∪ D2 ) . The energy balance inside the solids is  D1 ∪D2



[e]d Ω =

Σ1 : D(

D1

+ − +

 D2

Σ2 : D(



∂ D1



D1

− U+ 1 + U1 ) + B p1 [β1 ] + H p1 · grad[β1 ] d Ω1 2 − U+ 2 + U2 ) + B p2 [β2 ] + H p2 · grad [β2 ] d Ω2 2

Σ (T1 Q p1 · N1 )dΓ1 −

Σ (T1 B1 )d Ω1 +

 D2



∂ D2

Σ (T2 Q p2 · N2 )dΓ2

Σ (T2 B2 )d Ω 2 ,

where the subdomains D1 and D2 are interior to each solid and where T1 Q p1 , T2 Q p2 are the heat flux vectors in Ω 1 and in Ω2 . With this relationship, we get  O1

[es1 ]dΓ1 +

 O2

[es2 ]dΓ2 =

 O

+ + −

R p · Ds (

 ∂ D1



∂ D2



O

U+ + U− )d Γ 2

  Σ T1 (Q p1 · N1 + Q p1 + Bs1 ) d Γ1   Σ T2 (Q p2 · N2 + Q p2 + Bs2 ) d Γ2

Σ (T1 Q p1 + T2 Q p2 ) dΓ .

This results in the energy balance in Ω 1 [e1 ] + div Σ (T1 Q p1 ) = Σ1 : D(

− U+ 1 + U1 ) + B p1 [β1 ] + H p1 · grad [β1 ] + Σ (T1 B1 ) 2

− U+ 1 + U1 ) + B p1 [β1 ] + H p1 · grad [β1 ] + T 1 Σ (B1 ) 2 + [T1 ] Δ (B1 ),

= Σ1 : D(

or [e1 ] + div(T 1 (Σ (Q p1 ))) + div([T1 ] Δ (Q p1 )) = [e1 ] + T 1 div(Σ (Q p1 )) + gradT 1 · Σ (Q p1 ) + [T1 ] div Δ (Q p1 ) + grad[T1 ] · Δ (Q p1 )

198

11 Collisions of Deformable Bodies and Phase Change

= Σ1 : D(

− U+ 1 + U1 ) + B p1 [β1 ] + H p1 · grad[β1 ] + T 1 Σ (B1 ) + [T1 ] Δ (B1 ), 2

where we denote as usual Δ (X) = (X + − X − )/2, Σ (X) = X + + X − and we use the relationship Σ (XY ) = Σ (X)Y + Δ (X)[Y ], with Y = (Y + +Y − )/2 and [Y ] = Y + −Y − . We choose the volume free energies as

Ψ1 = −C1 T1 ln T 1 + Ψˇ1 , with

k1 L1 Ψˇ1 (T1 , β1 , grad β1 ) = −β1 (T1 − T0 ) + (grad β1 )2 + I(β1 ). T0 2

We have   + + − ˇ − − Ψˇ1 = Ψˇ1 (T + 1 , β 1 , grad β1 ) − Ψ1 (T 1 , β 1 , grad β1 ) + + − ˇ + − = Ψˇ1 (T + 1 , β 1 , grad β1 ) − Ψ1 (T1 , β1 , grad β1 ) − − +Ψˇ1 (T1+ , β1− , grad β1− ) − Ψˇ1(T − 1 , β 1 , grad β1 ),

and

L1 Ψˇ1 (T1+ , β1− ) − Ψˇ1(T1− , β1− ) = −β1− [T1 ] . T0

Thus L1 [Ψ1 ] = [−C1 T1 ln T1 ] + Ψˇ1 (T1+ , β1+ ) − Ψˇ1 (T1+ , β1− ) − β1− [T1 ] T0 ≤ [−C1 T1 ln T1 ] − β1−

L1 fe fe [T1 ] + B p1 [β1 ] + H p1 · grad[β1 ] , T0

with



L1 fe fe B p1, H p1 ∈ ∂ Ψˇ1 (T1+ , β1+ , grad β1+ ) = − (T1+ − T0 ) + ∂ I(β1+ ), k1 grad β1+ , T0 (11.3) where the subdifferential is computed with respect to β1+ and grad β1+ . Moreover [e1 ] = [Ψ1 + T1 s1 ] = [Ψ1 ] + s1 [T1 ] + T 1 [s1 ] fe fe ≤ −S1 [T1 ] + T 1 [s1 ] + B p1 [β1 ] + H p1 · grad[β1 ] ,

by denoting [−C1 T1 ln T1 ] − β1−

L1 [T1 ] + s1 [T1 ] = −S1 [T1 ] . T0

11.2 The First Law of Thermodynamics

199

Thus the energy balance in Ω 1 is T 1 [s1 ] + T 1 div(Σ (Q p1 )) ≥ Σ1 : D

− U+ 1 + U1 2



fe fe + (B p1 − B p1 ) [β1 ] + (H p1 − H p1 ) · grad [β1 ] + T 1 Σ (B1 )

+ [T1 ] (Δ (B1 ) + S1 − div Δ (Q p1 )) − grad[T1 ] · Δ (Q p1 ) − gradT 1 · Σ (Q p1 ). When the temperature discontinuity is zero, S1 is zero since

∂Ψ + s = 0. ∂T This relationship is extended by assuming that there is no dissipation with respect to [T1 ] and grad [T1 ]. Then

Δ (B1 ) + S1 − div Δ (Q p1 ) = 0, Δ (Q p1 ) = 0, which give − Δ (B1 ) + S1 = 0, Q+ p1 = Q p1 = Q p1 .

(11.4)

The energy balance becomes [s1 ] + 2 divQ p1 ≥

1 T1



 Σ1 : D

− U+ 1 + U1 2



+ (B p1 − B p1) [β1 ] + (H p1 − H p1 ) · grad[β1 ] fe

−

fe

2 grad T 1 · Q p1 + Σ (B1 ), T1

(11.5)

In the same way, the energy balance in Ω2 is [s2 ] + 2 divQ p2

1 ≥ T2





− U+ 2 + U2 Σ2 : D 2



fe fe + (B p2 − B p2 ) [β2 ] + (H p2 − H p2 ) · grad[β2 ]

−

2 grad T 2 · Q p2 + Σ (B2 ), T2

(11.6)

with − Δ (B2 ) + S2 = 0, Q+ p2 = Q p2 = Q p2 .

(11.7)

200

11 Collisions of Deformable Bodies and Phase Change

Remark 11.2. In this section the discontinuity of the free energy [Ψ ] = Ψ (T + , β + , grad β + ) − Ψ (T − , β − , grad β − ) = Ψ (T + , β + , grad β + ) − Ψ (T + , β − , grad β − ) +Ψ (T + , β − , grad β − ) − Ψ (T − , β − , grad β − ), is as in Sect. 3.6, split into two terms Ψ (T + , β + , grad β + ) − Ψ (T + , β − , grad β − ) and Ψ (T + , β − , grad β − ) − Ψ (T − , β − , grad β − ). The first one is treated as in Sect. 3.6 by using the convexity of function (β , grad β ) → Ψ (T + , β , grad β ). The second one is replaced by

[−C1 T1 ln T1 ] L1 − β1− [T1 ] T0

[T1 ] ,

which is different from

L1 ∂ΨT (T1− , β − , grad β − ) [T1 ] = −C1 (1 + lnT1− ) − β1− [T1 ] . T0 This difference appears in (11.4) and (3.42) where the two S which are slightly different, give Δ (B) or the allocation of the external entropy production B+ + B− between temperatures T + and T − . We think that the two way of doing are convenient: to use either the differential quotient or the derivative of −C1 T1 ln T1 . This remark give a hint to the way to deal with non convex free energies: in case (β , grad β ) → Ψ (T + , β , grad β ) is not convex, the discontinuities may be replaced by differential quotients. But it is a long and complicated task. The energy balance on ∂ Ω 1 is   [es1 ] = T 1 2Q p1 · N1 + Σ T1 (Q p1 + Bs1 ) = T 1 2Q p1 · N1 + T 1 Σ (Q p1 + Bs1 ) + Δ (Q p1 + Bs1 ) [T1 ] .

(11.8)

Moreover due to the relationship es1 = Ψs1 + ss1 T1 [es1 ] = [Ψs1 ] + ss1 [T1 ] + T 1 [ss1 ] = −Ss1 [T1 ] + T 1 [ss1 ], by denoting

[Ψs1 ] + ss1 [T1 ] = −Ss1 [T1 ] .

Then the energy balance is   T 1 [ss1 ] = T 1 2Q p1 · N1 + T 1 Σ (Q p1 + Bs1 ) + Δ (Q p1 + Bs1 ) + Ss1 [T1 ] .

11.2 The First Law of Thermodynamics

201

Assuming as usual no dissipation with respect to [T1 ]

Δ (Q p1 + Bs1 ) + Ss1 = 0,

(11.9)

T 1 [ss1 ] = T 1 2Q p1 · N1 + T 1 Σ (Q p1 + Bs1 ) .

(11.10)

the energy balance becomes

The different heat sources which produce the evolution of the internal energy are thermal. There is no mechanical source in (11.10). The energy balance on ∂ Ω 2 is   [es2 ] = T 2 2Q p2 · N2 + Σ T2 (Q p2 + Bs2 ) .

(11.11)

T 2 [ss2 ] = T 2 2Q p2 · N2 + T 2 Σ (Q p2 + Bs2 ) ,

(11.12)

Then

with no dissipation with respect to [T2 ]

Δ (Q p2 + Bs2 ) + Ss2 = 0, where Ss2 is defined by

(11.13)

[Ψs2 ] + ss2 [T2 ] = −Ss2 [T2 ] .

The energy balance on ∂ Ω 1 ∩ ∂ Ω 2 is [es1 ] + [es2] = R p · Ds

U+ + U− 2



+Σ (T1 Bs1 ) + Σ {T2 Bs2 } + T 1 2Q p1 · N1 +T 2 2Q p2 · N2 .

(11.14)

We have [es1 ] + [es2] = R p · Ds

U+ + U− 2



+ [T1 ] Δ (Bs1 ) + T 1 Σ (Bs1 ) + [T2 ] Δ (Bs2 ) +T 2 Σ (Bs2 ) +T 1 2Q p1 · N1 + T 2 2Q p2 · N2 .

202

11 Collisions of Deformable Bodies and Phase Change

By subtracting relationships (11.8) and (11.11), it results U+ + U− ) − [T1 ] Δ (Q p1 ) − T 1 Σ (Q p1 ) − [T2 ] Δ (Q p2 ) − T 2 Σ (Q p2 ) 2 U+ + U− δT ) − [T1 ] Δ (Q p1 ) − [T2 ] Δ (Q p2 ) − (Σ (Q p2 ) − Σ (Q p1) = R p · Ds ( 2 2 −T m (Σ (Q p2 ) + Σ (Q p1 ).

0 = R p · Ds (

Assuming as usual no dissipation with respect to the [T ]’s, we let

Δ (Q p1 ) = 0, Δ (Q p2 ) = 0.

(11.15)

− + − Then Q+ p1 = Q p1 = Q p1 , Q p2 = Q p2 = Q p2 and

0 = R p · Ds (

U+ + U− ) − δ T (Q p2 − Q p1 ) − 2T m (Q p2 + Q p1). 2

(11.16)

Remark 11.3. On ∂ Ω 1 ∩ ∂ Ω 2 , the relationships (11.9) and (11.13) become

Δ (Bs1 ) + Ss1 = 0, Δ (Bs2 ) + Ss2 = 0, due to (11.15).

11.3 The Second Law of Thermodynamics Surface entropies ss are introduced. The second law of thermodynamics is  D1 ∪D2

[s]d Ω +

 O1

[ss1 ]dΓ1 +

 O2

[ss2 ]d Γ2 ≥ − − + + + +

 ∂ D1 \O1



∂ D2 \O2



D1

O1 \O



O2 \O

O

2Q p2 · N2 d Γ2

Σ (B1 )d Ω1 +





2Q p1 · N1 d Γ1

 D2

Σ (B2 )d Ω2

Σ (Q p1 ) + Σ (Bs1 )dΓ1 Σ (Q p2 ) + Σ (Bs2 )dΓ2

Σ (Bs1 )dΓ1 +

 O

Σ (Bs2 )dΓ2 .

11.3 The Second Law of Thermodynamics

203

By integrating by parts  D1

[s1 ] + div(2Q p1 ) − Σ (Bs1 )d Ω1 +

+ ≥

 O1



O1 \O

+ +

[ss1 ]dΓ1 +

O2 \O

O

O2

D2

[s2 ] + div(2Q p2 ) − Σ (Bs2 )d Ω 2

[ss2 ]dΓ2

2Q p1 · N1 + Σ (Q p1 ) + Σ (Bs1 )dΓ1

 





2Q p2 · N2 + Σ (Q p2 ) + Σ (Bs2 )dΓ2

2Q p1 · N1 + Σ (Bs1 )dΓ1 +

 O

2Q p2 · N2 + Σ (Bs2 )dΓ2 ,

then  D1

[s1 ] + div(2Q p1 ) − Σ (Bs1 )d Ω1 +

+ ≥

 O1



 O1

+ +



O2

O

 O2

D2

[s2 ] + div(2Q p2 ) − Σ (Bs2 )d Ω 2

[ss2 ]dΓ2

 2Q p1 · N1 + Σ (Q p1 ) + Σ (Bs1 ) dΓ1

 

[ss1 ]dΓ1 +



 2Q p2 · N2 + Σ (Q p2 ) + Σ (Bs2 ) d Γ2

{−Σ (Q p1 ) − Σ (Q p2)}d Γ .

The preceding inequality gives [s1 ] + div(2Q p1 ) − Σ (B1 ) ≥ 0, in Ω1 ,

(11.17)

[s2 ] + div(2Q p2 ) − Σ (B2 ) ≥ 0, in Ω2 ,   [ss1 ] − 2Q p1 · N1 + Σ (Q p1 ) + Σ (Bs1 ) ≥ 0, on ∂ Ω 1 ,   [ss2 ] − 2Q p2 · N2 + Σ (Q p2 ) + Σ (Bs2 ) ≥ 0, on ∂ Ω 2 ,

(11.18) (11.19) (11.20)

[ss1 ] + [ss2 ]     − 2Q p1 · N1 + Σ (Bs1 ) − 2Q p2 · N2 + Σ (Bs2 ) ≥ 0, on ∂ Ω 1 ∩ ∂ Ω 2 . (11.21) The energy balances (11.10) and (11.12) imply the inequalities (11.19) and (11.20) are satisfied (they are equalities). It results that (11.21) is equivalent to

Σ (Q p1 ) + Σ (Q p2 ) = 2(Q p2 + Q p1) ≥ 0.

204

11 Collisions of Deformable Bodies and Phase Change

The energy balances (11.5), (11.6) and (11.16) induce to choose the basic assumptions to establish the constitutive laws: +

U1 + U− fe fe 1 Σ1 : D + (B p1 − B p1) [β1 ] + (H p1 − H p1 ) · grad[β1 ] 2 −2 gradT 1 · Q p1 ≥ 0, in Ω1 , (11.22) + −

U2 + U2 fe fe ) [β2 ] + (H p2 − H p2 ) · grad[β2 ] Σ2 : D + (B p2 − B p2 2 −2 gradT 2 · Q p2 ≥ 0, in Ω2 , +

U + U− R p · Ds − δ T (Q p2 − Q p1 ) ≥ 0, on ∂ Ω 1 ∩ ∂ Ω 2 . 2

(11.23) (11.24)

11.4 The Constitutive Laws We assume pseudopotential of dissipation +



U1 + U− ± 1 , [β1 ] , grad [β1 ] , grad T 1 , E1 , χ , = Φ1 D 2 +



U2 + U− ± ± ± 2 , [β2 ] , grad [β2 ] , grad T 2 , E2 , χ , Φ2 (E2 , Δ E2 , χ ) = Φ2 D 2

U+ + U− ± ± ± ), δ T , E , χ , Φs (δ E , E , χ ) = Φs Ds ( 2

Φ1 (Δ E1± , E1± , χ )

which imply relationships (11.22)–(11.24) are satisfied. The surface pseudopotential of dissipation Φs takes into account the impenetrability condition Ds (U+ ) · N1 ≥ 0, with the indicator function I+ (Ds (U+ ) · N1 ), (11.25) and we have

Φs (δ E ± , E ± , χ ) = Φs (δ E ± , E ± , χ ) + I+(Ds (U+ ) · N1 ), where quantities χ involve Ds (U− ).

11.4 The Constitutive Laws

205

The constitutive laws are fe

fe

(Σ1 , B p1 − B p1, H p1 − H p1 , −2Q p1 ) +



U1 + U− ± 1 , [β1 ] , grad [β1 ] , grad T 1 , E1 , χ , ∈ ∂ Φ1 D 2

(11.26)

fe fe , H p2 − H p2 , −2Q p2 ) (Σ2 , B p2 − B p2

+

U2 + U− ± 2 ∈ ∂ Φ2 D , [β2 ] , grad [β2 ] , grad T 2 , E2 , χ , 2 +



U + U− ± (R p , −(Q p2 − Q p1 )) ∈ ∂ Φs Ds ,δT ,E ,χ , 2

(11.27)

where the subdifferential are computed with respect to the first quantities and not with respect to the E’s and χ . For example, we choose

+

U1 + U− ± 1 , [β1 ] , grad [β1 ] , grad T 1 , E1 , χ Φ1 D 2

+

U1 + U− 1 , [ β1 ] , grad [β1 ] , grad T 1 , T 1 = Φ1 D 2 +

2 U1 + U− k c 1 + (grad T 1 )2 + [β1 ]2 = 2kM D 2 2T 1 2 +

km (grad[β1 ])2 , 2

and



+

+

U + U− U + U− , δ T , E ± , χ = Φs Ds , δ T , T 1 , T 2 , Ds (U+ Φs Ds 2 2 +

2 d T + T 2 k ps U + U− + 1 (δ T )2 + I+(Ds (U+ ) · N1 ), = kls Ds 2 T 1T 2 2 where the k’s parameters and c denote different thermal conductivities and dissipative parameters. For the sake of simplicity, we assume Φ1 ≡ Φ2 . The pseudopotential of dissipation have been chosen in such a way that the constitutive laws are linear besides the compulsory impenetrability condition. We think that the basic physical phenomena are to be described by these laws. It is only when dealing with particular problems that sophisticated non linear constitutive laws are to be introduced. Remark 11.4. The coefficient of (δ T )2 in pseudopotential Φs is T 1 + T 2 kdps , T 1T 2 2

206

11 Collisions of Deformable Bodies and Phase Change

whereas it is

λ , 4 in pseudopotential Φ of Chap. 10 for collisions of rigid balls. In the small perturbation assumption where the temperature have values close to T0 , we have λ=

4kdps , T0

if the rigid body collision is understood as an approximation of the deformable body collision. The coefficient in front of kdps is useful to have rather simple boundary conditions (for instance, (11.51)). The constitutive laws are   − Σ1 = 2kM D U+ 1 + U1 , B p1 − B p1 = c [β1 ] , H p1 − H p1 = km grad [β1 ] , fe

−Q p1 =

fe

k grad T 1 , in Ω1 , T1

  − Σ2 = 2kM D U+ 2 + U2 , B p2 − B p2 = c [β2 ] , H p2 − H p2 = km grad [β2 ] , fe

−Q p2 =

fe

k grad T 2 , in Ω2 , T2

and   + − − dr R p = kls U+ 2 − U1 + U2 − U1 + P N1 , −(Q p2 − Q p1) =

T1 +T2 d k δ T , on ∂ Ω 1 ∩ ∂ Ω2 , T 1 T 2 ps

(11.28)

where Pdr is the impenetrability reaction + Pdr ∈ ∂ I+ (Ds (U+ ) · N1 ) = ∂ I+ ((U+ 2 − U1 ) · N1 ).

Remark 11.5. The pseudopotentials Φi does not have constraint on [βi ], because βi+ is between 0 and 1 due to the definition of Bnd i (see relationship (11.3) which involves ∂ I(β1+ )). Note that the constitutive laws are only dissipative as are the constitutive laws for rigid bodies collisions. This property is characteristic of collisions where even the

11.5 Evolution in a Collision

207

impenetrability reactions are dissipative in contrast with many other cases where the reactions to internal constraints are workless (see Sect. 3.4.1). The following theorem ensures that the second law of thermodynamics is satisfied with the constitutive laws that have just been chosen. Theorem 11.1. If the constitutive laws are satisfied and if the temperatures are positive, then: • The impenetrability internal constraint is satisfied. • The second law of thermodynamics is satisfied. Proof. The proof is straightforward.



11.5 Evolution in a Collision The data of this problem are the state and the velocities before the collision, as well as the exterior actions: the percussions F p , T p , A, a, and the heat impulse sources Σ (T B) in Ω1 and Ω 2 , Σ T (Q p ) on ∂ Ω1 \(∂ Ω 1 ∩ ∂ Ω2 ) and ∂ Ω2 \(∂ Ω1 ∩ ∂ Ω2 ), and Σ T (Bs ) in ∂ Ω1 and ∂ Ω2 . The unknowns are the state quantities (T, β ) and the velocities after the collision. The mechanical equations are

ρ1 [U1 ] = div Σ1 + F p1 , 0 = div H p1 − B p1 + A1 , in Ω1 ,

(11.29)

ρ2 [U2 ] = div Σ2 + F p2 , 0 = div H p2 − B p2 + A2 , in Ω2 ,

(11.30)

Σ1 N1 = R p + T p1 , H p1 · N1 = a1 , Σ2 N2 = −R p + T p2 , H p2 · N2 = a2 , on ∂ Ω1 ∩ ∂ Ω 2 ,

(11.31)

Σ1 N1 = T p1 , H p1 · N1 = a1 , on ∂ Ω1 \(∂ Ω1 ∩ ∂ Ω2 ),

(11.32)

Σ2 N2 = T p2 , H p2 · N2 = a2 , on∂ Ω2 \(∂ Ω1 ∩ ∂ Ω2 ).

(11.33)

208

11 Collisions of Deformable Bodies and Phase Change

The thermal equations are [e1 ] + div(2T 1 Q p1 ) = Σ1 : D(

− U+ 1 + U1 ) + B p1 [β1 ] + H p1 · grad[β1 ] + Σ (T1 B1 ), in Ω 1 , 2 (11.34)

[e2 ] + div(2T 2 Q p2 ) = Σ2 : D(

− U+ 2 + U2 ) + B p2 [β2 ] + H p2 · grad[β2 ] + Σ (T2 B2 ), in Ω 2 , 2 (11.35)

[es1 ] = T 1 2Q p1 · N1 + Σ {T1 (Q p1 + Bs1 )}, on ∂ Ω1 , Remark 11.6. The last relationship and the two following ones are boundary conditions for the partial differential equations which are to be satisfied by the temperature in Ω1 and in Ω2 . [es2 ] = T 2 2Q p2 · N2 + Σ {T2 (Q p2 + Bs2 )}, on ∂ Ω2 , U+ + U− ) + Σ {T2 Bs2 } + Σ {T1Bs1 } 2 +2T 1 Q p1 · N1 + 2T 2 Q p2 · N2 , on ∂ Ω1 ∩ ∂ Ω2 .

[es1 ] + [es2] = R p · Ds (

The equations are completed by the relationships

Δ (B1 ) + S1 = 0, in Ω1 , Δ (B2 ) + S2 = 0, in Ω2 ,  Δ Q p1 + Bs1 + Ss1 = 0, on ∂ Ω1 , 

Δ (Q p2 + Bs2 ) + Ss2 = 0, on ∂ Ω2 , Δ (Q p1 ) = 0, Δ (Q p2 ) = 0, on ∂ Ω 1 ∩ ∂ Ω2 . Equation (11.4) with the data Σ (T1 B1 ) give the two heat impulses B1+ and in Ω1 , depending on the temperatures T1+ and T1− . With the data Σ (T1 Q p1 ), Σ (T1 Bs1 ), and (11.8), the two heat impulses B1+ and B1− give the heat flux T 1 2Q p1 · N1 on ∂ Ω1 \(∂ Ω1 ∩ ∂ Ω 2 ) which is the boundary condition for (11.34). Equations (11.9), (11.13), (11.8), (11.11), (11.14), the constitutive laws (11.28) and the data Σ (T1 Bs1 ), Σ (T2 Bs2 ) give the Bs , Q p , and the two heat fluxes T 1 2Q p1 · N1 and T 2 2Q p2 · N2 on ∂ Ω1 ∩ ∂ Ω2 depending on the temperatures and the mechanical works. The heat fluxes T 1 2Q p1 ·N1 and T 2 2Q p2 ·N2 on ∂ Ω 1 ∩ ∂ Ω2 are the boundary conditions on ∂ Ω1 ∩ ∂ Ω 2 for (11.34) and (11.35). Thus the equations provide the boundary conditions for solving the two (11.34) and (11.35) which give the B1−

11.5 Evolution in a Collision

209

two future temperatures T1+ and T2+ . In the following sections examples of those mechanical and thermal equations are investigated.

11.5.1 The Mechanical Evolution When Decoupled from the Thermal Evolution The equations of motion are (11.29)–(11.33). Let us recall that we have assumed that all the mechanical quantities are independent of the temperature and that the constitutive laws have been chosen simple in order to emphasize the properties of the equations   − Σ1 = 2kM D U+ 1 + U1 , in Ω 1 ,   − Σ2 = 2kM D U+ 2 + U2 , in Ω 2 , the positive parameter kM describes the dissipative properties of the material in a collision;   + − − dr R p = kls U+ 2 − U1 + U2 − U1 + P N1 , on ∂ Ω 1 ∩ ∂ Ω 2 , the positive parameter kls describes the dissipative properties of the contact surface in a collision. The properties of non-interpenetration reaction Pdr result from its definition + Pdr ∈ ∂ I+ (Ds (U+ ) · N1 ) = ∂ I− ((U+ 2 − U1 ) · N1 ). Introducing the constitutive laws in the equations of motion, it results by denoting Δ V the laplacian of vector V   + + − ρ 1 U+ 1 = kM graddiv U1 + Δ U1 + F p1 + ρ1 U1   − +kM grad div U− 1 + Δ U1 , in Ω 1 ,   + + − ρ 2 U+ 2 = kM graddiv U2 + Δ U2 + F p2 + ρ2 U2   − +kM grad div U− 2 + Δ U2 , in Ω 2 ,   + − − dr Σ1 N1 = kls U+ 2 − U1 + U2 − U1 + P N1 , Σ2 N2   + − − dr = −kls U+ 2 − U1 + U2 − U1 + P N2 , on ∂ Ω 1 ∩ ∂ Ω 2 ,

(11.38)

Σ1 N1 = T p1 , on ∂ Ω1 \(∂ Ω 1 ∩ ∂ Ω 2 ),

(11.39)

(11.36)

(11.37)

and Σ2 N2 = T p2 , on ∂ Ω 2 \(∂ Ω 1 ∩ ∂ Ω 2 ).

(11.40)

210

11 Collisions of Deformable Bodies and Phase Change

The preceding equations (11.36)–(11.40) have a variational formulation. Let C be the convex set of the kinematically admissible future velocities C = {(V1 , V2 ) ∈ V1 × V2 |(V2 − V1 ) · N1 ≥ 0, on ∂ Ω 1 ∩ ∂ Ω 2 }. It is easy to show that a solution of (11.36)–(11.40) satisfies + (U+ 1 , U2 ) ∈ C, ∀(V1 , V2 ) ∈ C, + + + + + + + a1 (U+ 1 , V1 − U1 ) + a2(U2 , V2 − U2 ) + b(U2 − U1 , V2 − V1 − (U2 − U1 )) + − − + + ≥ L1 (V1 − U+ 1 ) + L2 (V2 − U2 ) − b(U2 − U1 , V2 − V1 − (U2 − U1 )),

(11.41) with a1 (U, V) = b(U, V) = L1 (V) =



2kD(U) : D(V)d Ω1 +

Ω1



∂ Ω1 ∩∂ Ω 2



Ω1

−

Ω1

Ω1

ρ1 U · Vd Ω1 ,

kls U · V)d Γ1 ,

F p1 · Vd Ω1 +





 ∂ Ω1 \(∂ Ω1 ∩∂ Ω 2 )

2kD(U− 1 ) : D(V)d Ω 1 +

T p1 · Vd Γ1



Ω1

ρ1 U− 1 · Vd Ω 1 .

The functions a2 and L2 are defined in the same way. Problem (11.41) is a classical variational inequality which has a unique solution in a convenient functional framework [105,158,181,194]. Thus the future velocities U+ are uniquely determined by the past.

11.5.2 An Example: Collision of a Bar with a Rigid Support Consider the bar shown in Fig. 11.1 falling on a rigid fixed support (U2 = 0), [83]. The bar is assumed to be elastic when evolving smoothly (with Young modulus 2.5 1010 Pa, Poisson coefficient 0.3 and density 7877 kg /m3 ). When colliding with the rigid support, it is dissipative. The bar equation of motion when colliding the support is + + − − − ρ 1 U+ 1 = kM grad div U1 + kM Δ U1 + ρ1 U1 + kM graddiv U1 + kM Δ U1 ,

in the bar

  − dr Σ1 N1 = −kls U+ 1 + U1 − P N1 ,

11.5 Evolution in a Collision

211

Fig. 11.1 A bar falls with a vertical velocity of −1 m / s on a rigid support with an angle θ = 54◦ with the horizontal. Its length is 0.6 m. Its width is 0.012 m. The two ends are rounded. The velocities after the first collisions are shown

on the contact surface, and Σ1 N1 = 0, outside the contact surface. The dissipative parameters are chosen as: kM = 105 N s2 /m2 and kls = 8 106 N s2 / m3 . The velocities of the bar after the first collision where contact is not maintained, are not rigid body velocities. The bar vibrates afterwards. There is a competition between the slow rising motion after the collision and the fast vibrating motion resulting from the elasticity of the bar. The first oscillations are such that the bar which has not risen far above the support collides with it other times, [86]. The vertical position of a point of the lower part of the bar versus the time is shown in Fig. 11.2. Experiments have shown this behaviour, [201].

11.5.3 Thermal Evolution When the Mechanical Equations Are Decoupled from the Thermal Equations The equations to get the future temperatures and volume fractions are,

ρ1 [e1 ] + div(2T 1 Q p1 ) = Σ1 : D(

− U+ 1 + U1 ) + B p1 [β1 ] + H p1 · grad[β1 ] + Σ (T1 B1 ), 2

212

11 Collisions of Deformable Bodies and Phase Change

5

x 10–5

4.5 4

metres

3.5 3 2.5 2 1.5 1 0.5 0

0

1

2

3

4 secondes

5

6

7

8 x 10–4

Fig. 11.2 The vertical position of a point of the lower part of the bar which collides with the rigid support. The bar vibrates after the first collision. It results microrebounds before the bar rises sufficiently above the rigid support [83]

0 = div H p1 − B p1 + A1, in Ω1 ,

ρ2 [e2 ] + div(2T 2 Q p2 ) − U+ 2 + U2 ) + B p2 [β2 ] + H p2 · grad[β2 ] + Σ (T2 B2 ), 2 0 = div H p2 − B p2 + A2, in Ω2 ,   [es1 ] = T 1 2Q p1 · N1 + Σ T1 (Q p1 + Bs1 ) ,

= Σ2 : D(

H p1 · N1 = a1 , on ∂ Ω 1 \(∂ Ω 1 ∩ ∂ Ω 2 ),   [es2 ] = T 2 2Q p2 · N2 + Σ T2 (Q p2 + Bs2 ) , H p2 · N2 = a2 , on ∂ Ω 2 \(∂ Ω 1 ∩ ∂ Ω 2 ), by using (11.8), (11.11), and (11.14) together with the constitutive law (11.28) [es1 ] =

T1 U+ + U− ) R p · Ds ( T1 +T2 2

2T 1 T 2 (Q p2 − Q p1) + Σ {T1 Bs1 }, T1 +T2

T2 U+ + U− R p · Ds ( [es2 ] = ) T1 +T2 2 +T 1 2Q p1 · N1 −

+T 2 2Q p2 · N2 +

(11.42)

2T 1 T 2 (Q p2 − Q p1) + Σ {T2 Bs2 }, on ∂ Ω 1 ∩ ∂ Ω 2 , T1 +T2 (11.43)

11.5 Evolution in a Collision

and

213

H p1 · N1 = a1 , H p2 · N2 = a2 , on ∂ Ω 1 ∩ ∂ Ω 2 .

The equations are completed by relationships (11.4), (11.7), (11.9), (11.13) and, (11.15). The constitutive laws have already been chosen, (see formulas (11.28)). Let us recall those we need − Q p1 = −(Q p2 − Q p1) =

k k grad T 1 , −Q p2 = grad T 2 , T1 T2 T1 +T2 d k δT, T 1 T 2 ps

fe + c [β1 ] = − B p1 ∈ B p1

L1 + (T − T0 ) + ∂ I(β1+ ) + c [β1 ] , T0 1

fe H p1 = H p1 + km grad [β1 ] , fe

B p2 ∈ B p2 + c [β2 ] = −

L2 + (T − T0 ) + ∂ I(β2+ ) + c [β2 ] , T0 2

fe + km grad [β2 ] , H p2 = H p2

where the constants k and kdps denote the different thermal conductivities and c and km are dissipative parameters. We have already chosen the volume free energies

Ψ (T, β ) = −Ci T ln T − β

Li ki (T − T0 ) + (grad β )2 + I(β ). T0 2

Since we assume the mechanical properties do not depend on the temperature, the surface free energies have a unique temperature dependent term which is chosen simple Ψs (T ) = −Cs T ln T, where the C’s are heat capacities. We assume that the exterior heat impulses are null: Σ (T1 B1 ) = Σ (T2 B2 ) = 0, Σ (T1 Q p1 ) = 0 on ∂ Ω1 \(∂ Ω1 ∩ ∂ Ω 2 ), Σ (T1 Bs1 ) = 0 on ∂ Ω 1 , Σ (T2 Q p2 ) = 0 on ∂ Ω2 \(∂ Ω1 ∩ ∂ Ω2 ), Σ (T2 Bs2 ) = 0 on ∂ Ω2 and that no external percussion work is provided A1 = A2 = 0, and a1 = a2 = 0. Remark 11.7. It is also possible to assume that the exterior entropy impulses are given. The energy balances (11.34), (11.35), (11.8), (11.40) and (11.42), (11.43), with the constitutive laws give the equations C1 [T1 ] + L1 [β1 ] − 2kΔ T 1 = Σ1 : D(

− U+ 1 + U1 ) + B p1 [β1 ] + H p1 · grad[β1 ] , 2

(11.44)

214

11 Collisions of Deformable Bodies and Phase Change

c [β1 ] − km Δ [β1 ] + ∂ I(β1+ )

L1 + (T − T0 ), in Ω1 , T0 1

(11.45)

C2 [T2 ] + L2 [β2 ] − 2kΔ T 2 − U+ 2 + U2 ) + B p2 [β2 ] + H p2 · grad[β2 ] , 2 L2 c [β2 ] − km Δ [β2 ] + ∂ I(β2+ ) (T2+ − T0 ), in Ω2 , T0

= Σ2 : D(

(11.46) (11.47)

where Δ denotes the laplacian operator, C1s [T1 ] + 2k

∂T1 ∂ [β1 ] = 0, km = 0, on ∂ Ω1 \(∂ Ω 1 ∩ ∂ Ω 2 ), ∂ N1 ∂ N1

(11.48)

C2s [T2 ] + 2k

∂T2 ∂ [β2 ] = 0, km = 0, on ∂ Ω2 \(∂ Ω 1 ∩ ∂ Ω 2 ), ∂ N2 ∂ N2

(11.49)

these relationships are the boundary conditions on ∂ Ω 1 \(∂ Ω1 ∩ ∂ Ω2 ) and ∂ Ω2 \(∂ Ω1 ∩ ∂ Ω2 ) of (11.44) and (11.46) in Ω1 and Ω2 , +

U + U− R p · Ds + 2kdps δ T , (11.50) 2

+ T2 ∂T U + U− − 2kdps δ T , (11.51) C2s [T2 ] + 2k 2 = R p · Ds ∂ N2 T1 +T2 2

T1 ∂T C1s [T1 ] + 2k 1 = ∂ N1 T1 +T2

km

∂ [β1 ] ∂ [ β2 ] = 0, km = 0, on ∂ Ω 1 ∩ ∂ Ω 2 . ∂ N1 ∂ N2

As already mentioned, relationships (11.50) and (11.51) together with (11.42) and (11.43) give the boundary conditions on ∂ Ω1 ∩ ∂ Ω2 for the two partial differential equations (11.44) and (11.46) equations in Ω1 and Ω2 giving the future temperature T1+ and T2+ . The equations giving the temperatures and the volume fractions are coupled due to the phase change latent heat.

11.5.4 The Temperature Variation in a Collision Because collisions are dissipative, their thermal effect should be to warm the colliding solids. In this section, we investigate this problem. Let us assume that the temperatures before the collision are uniform (they do not depend on x) in each solid and lower than the phase change temperature, i.e., T1− < T0 , T2− < T0 . We assume also that β1− = β2− = 0. The effect of kdps is to equalize the temperature on the boundary thus we assume that kdps = 0 in order to avoid this effect. We assume also that H p1 = H p2 = 0. By using the relationship T = T − + [T ]/2, it results the

11.5 Evolution in a Collision

215

equations, C1 [T1 ] + L1 [β1 ] − kΔ [T1 ] = Σ1 : D( c [β1 ] + ∂ I(β1+ )

L1 + (T − T0 ), in Ω1 , T0 1

C2 [T2 ] + L2 [β2 ] − kΔ [T2 ] = Σ2 : D( c [β2 ] + ∂ I(β2+ )

− U+ 1 + U1 ) = Y1 , 2

− U+ 2 + U2 ) = Y2 , 2

L2 + (T − T0 ), in Ω2 , T0 2

(11.52)

because we have equations of motion B p1 = 0 and B p2 = 0, C1s [T1 ] + k

∂ [T1 ] = 0 = Y1s , on ∂ Ω 1 \(∂ Ω1 ∩ ∂ Ω2 ), ∂ N1

C2s [T2 ] + k

∂ [T2 ] = 0 = Y2s , on ∂ Ω 2 \(∂ Ω1 ∩ ∂ Ω2 ), ∂ N2

and

U+ + U− R p · Ds ( ) = Y1s , 2

T2 U+ + U− ∂ [T2 ] = R p · Ds ( ) = Y2s , on ∂ Ω 1 ∩ ∂ Ω 2 . C2s [T2 ] + k ∂ N2 T1 + T2 2

∂ [T1 ] T1 C1s [T1 ] + k = ∂ N1 T1 + T2

Let us define X1 = np([T1 ]) and X2 = np([T2 ]), where np(A) is the negative part of A, np(A) = sup{−A, 0}. The equation of motion for β1 is fe 0 = B p1 ∈ B p1 + ∂ Φ1 ([β1 ]) = −

=−

L1 + (T − T0 ) + ∂ I(β1+ ) + ∂ Φ1 ([β1 ]) T0 1 L1 + (T − T0 ) + ∂ I([β1 ]) + ∂ Φ1 ([β1 ]), T0 1

because β1+ = [β1 ]. We get L1 + (T − T0 ) [β1 ] ∈ (∂ I([β1 ]) + ∂ Φ1 ([β1 ])) [β1 ] , T0 1 and L1 + (T − T0 ) [β1 ] ≥ 0. T0 1

(11.53)

216

11 Collisions of Deformable Bodies and Phase Change

It results that if [β1 ] > 0, (T1+ − T0 ) ≥ 0 and (T1+ − T1− ) ≥ (T1+ − T0 ) ≥ 0: then X1 = np([T1 ]) = 0. Thus we have proved that L1 [β1 ] X1 = 0,

(11.54)

because either [β1 ] or X1 is zero. By multiplying (11.44), (11.48), (11.42) by X1 = np([T1 ]) and (11.46), (11.49), (11.43) by X2 = np([T2 ]), using relationship (11.54), and integrating over the Ω ’s and ∂ Ω ’s − −

 

Ω1 Ω2

=

C1 (X1 )2 d Ω 1 − C2 (X2 )2 d Ω 2 −

 Ω1

Y1 X1 d Ω1 +

 Ω1





Ω2

k (grad X1 )2 d Ω1 − (k grad X2 )2 d Ω2 −

∂ Ω1

Y1s X1 d Γ1 +

 Ω2

 ∂ Ω1



∂ Ω2

C1s (X1 )2 d Γ1 C2s (X2 )2 d Γ2

Y2 X2 d Ω2 +

 ∂ Ω2

Y2s X2 d Γ2 . (11.55)

Due to the constitutive laws and properties of the subdifferentials, the right-hand sides Y ’s and Ys ’s of the previous equations (11.42)–(11.49) are non negative. Then the right-hand side of (11.55) is non negative. But the left-hand side is non positive. Thus both of them are zero and X1 = np([T1 ]) and X2 = np([T2 ]) are equal to zero. Thus [T1 ] ≥ 0 and [T2 ] ≥ 0. As one expects, the temperatures increase when two solids collide. Phase change can occur. Relationship (11.53) shows that if it occurs at point x, temperature Ti+ (x) at that point is larger than the phase change temperature T0 .

Chapter 12

Phase Change Depending on a State Quantity: Liquid–Vapor Phase Change

Water may be solid, liquid and gaseous. We investigate in this chapter the liquid vapor phase change. Compared to the solid liquid phase change, there is a major difference: the phase change temperature depends on a parameter, for instance on the pressure. It is well known that water boils at the Mont Blanc summit at a lower temperature than at sea level. Moreover, if the temperature is larger than Tc , the critical temperature, water is only in its gaseous phase. At temperature lower than critical temperature Tc , water may be either liquid or gaseous. Experiments show that the phase change temperature depends on the pressure as shown in Fig. 12.1. A complete physical description may be found in [140, 147, 209]. Experiments show also that all the physical properties depend on two parameters as shown in Fig. 12.1, for instance T and pressure p. Numerous abaci, called thermodynamical diagrams, may be built. For instance, if P and specific volume τ are chosen, isovalues of temperature T (P, τ ) are drawn in the τ , P frame producing the diagram shown in Fig. 12.2. The diagram τ , T , the Clapeyron diagram, is shown in Fig. 12.3 with isovalues of pressure P(τ , T ). These diagrams are obtained with the two phases being in contact on a surface. This is a surfacic phase change as it occurs when water boils. But in a shower room after taking a shower, the mist is a mixture of air, of vapor which is transparent and of microscopic droplets of water which make the mixture not so transparent. The same mixture may be seen in fog and one may guess that clouds are also mixtures of air, vapor and liquid water (and even of ice which is assumed not to be present in the following predictive theories). These mixtures can be in contact with either liquid or pure vapor. These surfacic phase changes are also described in the diagrams. For instance, in Fig. 12.3, the vapor content βv is shown. Such mixtures, mist, fog or clouds, evolve in volumes. When a phase change occurs, it is a volumic phase change. In this chapter, we investigate the surfacic phase change and in the next chapter the volumic phase change. The results are similar.

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 12, © Springer-Verlag Berlin Heidelberg 2012

217

12 Phase Change Depending on a State Quantity: Liquid–Vapor Phase Change

Pressure atm

218

Critical point 374 °C, 218 atm

Solid

Vapor

Liquid

1

0.05

Triple point

0 0.01

100

Temperature °C

Fig. 12.1 Schematic pressure P versus temperature T for water in contact with vapor Fig. 12.2 A typical pressure P versus specific volume τ for different temperatures

P

C

Liquid

A

B

Vapor B’

Liquid+Vapor

A’ τ

12.1 The Temperature Is Lower than Critical Temperature Tc Let us consider liquid water in contact with vapor at an equilibrium, for instance boiling water. The contact surface, the phase change surface, is a free surface between a liquid domain with liquid volume fraction βl = 1 and a vapor domain with vapor volume fraction βv = 1 − βl = 1. On this surface [βl ] = [βv ] = 1. Applying the

12.1 The Temperature Is Lower than Critical Temperature Tc

T pc

Temperature

Fig. 12.3 The thermodynamical diagram temperature T versus specific volume τ for vapor in contact with liquid water. On the right of the green line there is vapor. On the left of the blue line there is liquid. In between the two lines there is a mixture of liquid and vapor. Some isopressure and isomixture curves are shown

219

C p1

τl

τc

τv

Specific volume τ

results of Sect. 3.5, relationships (3.34), (3.35) and (3.36) give  ] + s[T ] − T N [ 1 ]) = ∂ Φ , − ([Ψ ρ ∂m Q=

∂Φ , ∂ [T ]

TT =

∂Φ , ∂ [UT ]

H·N =

∂Φ , ∂ [β˙ ]

 = τΨ and s = τ s are the where density is ρ = 1/τ with specific volume τ , Ψ specific free energy and specific entropy whereas Ψ and s are the volumic free energy and volumic entropy. Following experiments, we assume no dissipation and no discontinuity of temperature on the phase change surface, and we choose surface pseudopotential of dissipation

Φ = I 0 ([T ]), where I0 is the indicator function of the origin of R. Because the two phases in contact are at an equilibrium, the velocity discontinuity [U] is null and we get from (3.28) [T] = 0,

(12.1)

the continuity of the stress which is the continuity of the pressure (T = σ N = −Pcloud N is the stress on the phase change surface, normal vector N is defined in Sect. 3.5). The definition of pressures (13.46) given below (see Sect. 13.11.2)

220

12 Phase Change Depending on a State Quantity: Liquid–Vapor Phase Change

Pcloud = pl

1 1 − Ψl = pv − Ψv , τl τv

(12.2)

with specific pressures pl and pv given by formula (13.38) in Remark 13.9 pl = −τl2

∂Ψl ∂Ψv , pv = −τv2 , ∂ τl ∂ τv

give the pressure in the cloud Pcloud = −τl

∂Ψl ∂Ψv − Ψl = −τv − Ψv . ∂ τl ∂ τv

(12.3)

The surface constitutive laws give  ] + Pcloud [ 1 ]) = 0, − ([Ψ ρ

(12.4)

[T ] = 0, Q ∈ R, TT = 0, H·N = 0 .

(12.5)

We get from relationships (12.2) and (12.4) p l = pv .

(12.6)

On the phase change surface there are three relationships, (12.3), (12.6), between the four parameters τl , τv , Pcloud and T . Thus they depend on one parameter. If we choose the volume fraction, we get the functions giving pressure Pcloud (τl ) (the B BC dotted line on the left) and Pcloud (τv ) (the CAA dotted line on the right) in Fig. 12.2. In the diagram of Fig. 12.3, the functions giving the temperature are T (τl ) (the blue curve) and T (τv ) (the green curve). If we choose temperature T , we get Pcloud (T ), the curve separating liquid and vapor domains in Fig. 12.1. The phase change depends on a parameter, for instance Pcloud or T . When Pcloud or T is chosen, all the physical quantities are known. Whereas for the solid liquid phase change, the ice water phase change, the phase change does not depend on any parameter: the phase change temperature is fixed. The last constitutive law, (12.5), prescribes that the work flux H · N is continuous. With constitutive law (13.26) for Hv and Hl , grad βv ·N and grad βl ·N are continuous but βv and βl are discontinuous. Remark 12.1. Formula (12.4) is analogous to the formula which gives the pressure in the volume as formula (13.46).

12.2 The Temperature Is Larger than Critical Temperature Tc

221

12.1.1 Partial Phase Change Let us consider liquid water (βl1 = 1) in contact with a mixture of vapor and liquid water (βl2 = 1). Due to relationship (12.1), the pressures on each side of the phase change surface are equal. Their value given by (13.46) is Pcloud = −τl

∂Ψl (T, τl ) − Ψl (T, τl ) ∂ τl

= −τl2

∂Ψl ∂Ψv (T, τl2 ) − Ψl (T, τl2 ) = −τv2 (T, τv2 ) − Ψv (T, τv2 ). ∂τ ∂τ (12.7)

It results from these equations that τl2 = τl1 = τl , τl being the common value. Relationship (12.4)  ] + Pcloud [ 1 ]) = 0, −([Ψ ρ gives with (12.7) 2 pv2 = −τv2

∂Ψv ∂Ψl (τv2 , T ) = pl = −τl2 (τl , T ) . ∂ τv2 ∂τ

(12.8)

The two relationships (12.7) and (12.8), which are identical to relationships (12.2) and (12.6), give τl = τl (T ) and τv2 = τv2 (T ). These relationships with mixture specific volume τ defined by 1 βv2 1 − βv2 = + . τ τv2 τl imply that T = T (τ , βv2 ). Such a relationship τ → T (τ , βv2 ) is shown in Fig. 12.3 for βv2 = 0, 0.5 and 1. Let us note that the results are the same for the contact of liquid water and vapor and for the contact of mixtures of liquid and vapor.

12.2 The Temperature Is Larger than Critical Temperature Tc Let us consider liquid water in contact with vapor at an equilibrium. The contact surface, the phase change surface, is a free surface between a liquid domain with liquid volume fraction βl = 1 and a vapor domain with vapor volume fraction βv = 1 − βl = 1. On this surface [βl ] = [βv ] = 1. Applying the results of previous section, we have

222

12 Phase Change Depending on a State Quantity: Liquid–Vapor Phase Change

pl pv − Ψl = − Ψv , τl τv

(12.9)

∂Ψl ∂Ψv = pv = −τv2 . ∂ τl ∂ τv

(12.10)

Pcloud = and pl = −τl2

When temperature T is larger than critical temperature Tc , the free energies are such that Ψl (τl , T ) = (τl , T ), Ψv (τv , T ) = (τv , T ), see Sect. 13.15 below. Relationship (12.10) gives −τl2

∂ ∂ (τl , T ) = −τv2 (τv , T ). ∂τ ∂τ

 ] = 0, which is Thus τl = τv and [τ ] = [1/ρ ] = 0. Relationship (12.4) gives [Ψ satisfied for any temperature. Thus at temperature larger than critical temperature Tc , it is impossible to have a difference between liquid and vapor. Thus there is no liquid vapor phase change. Volume fractions βl and βv are no longer state quantities.

Chapter 13

Clouds: Mixture of Air, Vapor and Liquid Water

Clouds are mixtures of air, vapor, liquid water and even of ice. For the sake of simplicity, we assume there is no ice phase. At each point of the mixture there is air and water either gaseous or liquid. Different indices are used: v for vapor, l for liquid, a for air, w for water. The clouds we see in the sky have a structure evolving with time. We think that the clouds cohesion results from local interactions: at a point x, the physical quantities, for instance, the velocities, the vapor content and liquid water content,.. do depend on their values in the neighbourhood of point x. A way to take into account these interactions is to introduce space derivatives which clearly depend on the neighbourhood values.

13.1 The Temperature Is Lower than Critical Temperature Tc When the temperature is lower than temperature Tc , called the critical temperature, vapor and liquid may coexist. In the sequel we assume T ≤ Tc up to Sect. 13.15 where the temperature is assumed to satisfy T ≥ Tc .

13.2 State Quantities The state quantities are: • The vapor, liquid and air specific volumes

τv , τl , τa .

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 13, © Springer-Verlag Berlin Heidelberg 2012

223

224

13 Clouds: Mixture of Air, Vapor and Liquid Water

• Air-vapor and liquid volume fractions

βv , βl . We introduce two volume fractions related by 1 = β v + βl , assuming that air and vapor coexist in the gaseous phase. It is necessary to have the two volume fractions because they have different velocities. • The gradients of the volume fractions grad βv , grad βl . They describe the influence of the volume fractions at a point onto their neighbourhood volume fractions. Such an influence is responsible for the cohesion of the clouds which is conspicuous when looking at them. • The temperature T, assuming the phases are in thermal equilibrium and have the same temperature. The non thermal equilibrium situation in case of fast phenomena, may be investigated, [97, 143]. We denote Ev = (τv , βv , βl , grad βv , T ), El = (τl , βv , βl , grad βl , T ), Ea = (τa , βv , T ), E = Ev ∪ El ∪ Ea .

13.3 Quantities Which Describe the Evolution and the Thermal Heterogeneity We denote Uv , Ul , Ua , the velocities of the three phases. The quantities which describe the evolution are the classical deformation velocities and other velocities accounting for different physical properties. These velocities may be called generalized deformation velocities.

13.3 Quantities Which Describe the Evolution and the Thermal Heterogeneity

225

13.3.1 Classical and Generalized Deformation Velocities They are: • The classical deformation velocities D(Uv ), D(Ul ), D(Ua ), where D is the usual strain rate operator. • The material derivatives of the volume fractions d l βv ∂ βv ∂ βv d v βv = + Uv · grad βv , = + Ul · grad βv , dt ∂t dt ∂t dv β l dl βl ∂ βl ∂ βl = + Ul · grad βl , = + Uv · grad βl , dt ∂t dt ∂t d a βv ∂ βv = + Ua · grad βv . dt ∂t • The gradients of the volume fractions velocities accounting for local interactions grad

dv β v d l βl , grad . dt dt

• The material derivatives of the specific volumes dv τv dl τl da τa , , . dt dt dt • The relative velocities of the phases Uv − Ul , Uv − Ua , Ul − Ua . These relative velocities may take into account dissipative interactions between the phases.

13.3.2 Thermal Heterogeneity It is described by gradT. We denote δ E the set of all these quantities: velocities of deformation and gradient of temperature

226

13 Clouds: Mixture of Air, Vapor and Liquid Water

δ E = (D(Uv ), D(Ul ), D(Ua ), dv β v d l βl dv βv dl βl da βv , , , grad , grad , dt dt dt dt dt Uv − Ul , Uv − Ua , Ul − Ua ,

(13.1)

dv τv dl τl da τa dl βv dv βl , , , , , dt dt dt dt dt grad T ). Velocities

dv τv dl τl da τa dl βv dv βl , , , , , dt dt dt dt dt do not appear in the power of the interior forces of the following section, they intervene like velocities d η /dt of Sect. 3.3.1.

13.4 Mass Balances Let us consider subdomains Ov (τ ) and Ol (τ ) of domain Ω (τ ) occupied by the mixture at time τ . The mass of vapor which is contained in Ov (τ ) is  Ov (τ )

βv (x, τ ) dΩ . τv (x, τ )

The mass of liquid water which is contained in Ol (τ ) is  Ol (τ )

βl (x, τ ) dΩ . τl (x, τ )

Let us choose Ov (τ ), Ol (τ ) such that at time t, Ov (t) = Ol (t) = O. The mass of water which is contained in subdomain O at time t is constant. It results

dv dτ





βv (x(τ ), τ ) dl d Ω (t) + dτ Ov (τ ) τv (x(τ ), τ )

∀t, ∀O, Ov (t) = Ol (t) = O,  βl (x(τ ), τ ) d Ω (t) = 0, Ol (τ ) τl (x(τ ), τ )



where the d v /dt and d l /dt are the material derivatives. The air mass balance is   βv (x(τ ), τ ) da d Ω (t) = 0. ∀t, ∀O, Oa (t) = O, dτ Oa (τ ) τa (x(τ ), τ )

13.5 Equations of Motion

227

By using relationship d dτ

 Ω (τ )

  f (x(τ ), τ )d Ω (t)=

 Ω (t)

 df (x,t)+ f (x,t) div U(x,t) d Ω , (13.2) dt

where material points x(t) move with velocity U, it results the mass balance equations: • For the water

βv βl d v βv d l βl ( ) + ( ) + div Uv + div Ul = 0. dt τv dt τl τv τl

(13.3)

• And for the air which occupies the same domain than the vapor da βv βv ( ) + div Ua = 0. dt τa τa

(13.4)

They are relationships between elements of δ E and they are going to be taken into account by the pseudopotential of dissipation.

13.5 Equations of Motion They result from the principle of virtual power. The important power is the power of the interior forces which introduces the different internal forces by their power.

13.5.1 The Power of the Interior Forces It is the sum of different powers, each of them being the scalar product of a velocity of deformation by the related internal force. The power we choose is Pint (Ω ) = −

 Ω

+Bvv

{σv : D(Uv ) + σl : D(Ul ) + σa : D(Ua ) dv β v dl βl da β v + Bll + Bva dt dt dt

d v βv d l βl + Hl · grad dt dt +fvl · (Uv − Ul ) + fva · (Uv − Ua ) + fla · (Ul − Ua )} d Ω ,

+Hv · grad

228

13 Clouds: Mixture of Air, Vapor and Liquid Water

where Ω is the domain occupied by the mixture at time t. Works B and works flux vectors H account for the microscopic motions which intervene in the vapor-liquid phase change. The novel quantity H describes the influence of the neighbourhood of a point onto this point for the microscopic motion (velocity dβ /dt represents the effect at the macroscopic level of the velocities at the microscopic level). Let us note, that in the same way, stress σ describes the effect of the neighbourhood of a point onto this point for the macroscopic motion. Forces fvl account for interactions, for instance frictions, between vapor and liquid with relative velocity Uv − Ul . Remark 13.1. Let us recall that the choice of the power of the internal forces depends on the scientist aiming to predict the motion of a mechanical system, [112]. The choice results mostly from the choice of δ E which contains the deformation velocities which are seen as being important and useful. The choice of the power of the internal forces is not fixed once for all. There are many possibilities resulting in different models. Remark 13.2. In terms of mathematics, an internal force is an element of the dual space of the linear space spanned by the related velocities of deformation. The two spaces are equipped with a bilinear form which is the power, see Appendix A. Let us recall that the quantity which is chosen by the engineer is the velocity of deformation. It is the quantity which is seen and measured in experiments. The internal forces are abstract quantities whose basic property is to be the dual quantity of the velocity of deformation in the bilinear form to give the power which is also measured in experiments. Let us define vector F mec = (σv , σl , σa , Bvv , Bll , Bva , Hv , Hl , fvl , fva , fla , 0, 0, 0, 0, 0, 0), and get Pint (Ω ) = −

 Ω

F mec · δ Ed Ω .

The last five but one 0 of vector F mec are related to velocities dη /dt which do not intervene in the power of the interior forces. The last 0 is related to gradT which do not either intervene in the power of the interior forces.

13.5.2 The Power of the Acceleration Forces The variation of the linear momentum of the vapor dv dt

 O

βv Uv d Ω τv

 =

 Ω (t)



 d v βv βv ( Uv ) + ( Uv ) div Uv d Ω , dt τv τv

13.5 Equations of Motion

229

using formula (13.2) introduces the acceleration. Thus the actual power of the acceleration forces is    v d βv βv ( Uv ) + ( Uv ) div Uv · Uv Pacc (Ω ) = dt τv τv Ω  l  d βl βl ( Ul ) + ( Ul ) div Ul · Ul + dt τl τl  a  d βa βa + ( Ua ) + ( Ua ) div Ua · Ua d Ω . dt τl τa The variation of the linear momentum

βv d v βv ( Uv ) + ( Uv ) div Uv , dt τv τv is the sum of the usual vapor linear momentum variation

βv d v Uv , τv dt and of the linear momentum variation resulting from the production of vapor 

 βv dv βv ( ) + div Uv Uv , dt τv τv

due to the phase change. Let us note that due to the air mass balance (13.4), we have 

βa da β a ( Ua ) + ( Ua ) div Ua dt τl τa

 =

βa da Ua . τl dt

13.5.3 Power of the Exterior Forces The actual power is Pext (Ω ) =

 Ω

+

ext ext fext v · Uv + fl · Ul + fa · Ua d Ω

 ∂Ω

ext ext gext v · Uv + gl · Ul + ga · Ua dΓ ,

ext where the fext v are the volume exterior forces and the gv are the surface exterior tractions, N is the normal outward vector to domain Ω .

230

13 Clouds: Mixture of Air, Vapor and Liquid Water

13.5.4 The Equations of Motion The equations of motion resulting from the principle of virtual power are: • The macroscopic linear momentum equations of motion. The linear momentum equations of motion

βv d v βv ( Uv ) + ( Uv ) div Uv = div σv − fvl − fva + fext v , in Ω , dt τv τv σv N = gext v , on ∂ Ω , dl β l βl ( Ul ) + ( Ul ) div Ul = div σl + fvl − fla + fext l , in Ω , dt τl τl

σl N = gext l , on ∂ Ω , β a da Ua = div σa + fla + fva + fext a , in Ω , τl dt σa N = gext a , on ∂ Ω ,

(13.5)

or

β v dv Uv + Uv τv dt



βl d l Ul + Ul τl dt

βv d v βv ( ) + div Uv dt τv τv



β dl βl ( ) + l div Ul dt τl τl

 = div σv − fvl − fva + fext v , in Ω , 

σv N = gext v on ∂ Ω , = div σl + fvl − fla + fext l , in Ω ,

σl N = gext l , on ∂ Ω , β a da Ua = div σa + fla + fva + fext a , in Ω , τl dt σa N = gext a , on ∂ Ω . Remark 13.3. Note that the sum of the two first equation is the water linear momentum equation of motion where the inertia term is   l  βv β l dl βl dv β v d βl ( ) + div Uv + Ul + Ul ( ) + div Ul dt τv τv τl dt dt τl τl  v  v l d βv βv d βl d βv = Uv + Ul + (Uv − Ul ) ( ) + div Uv , τv dt τl dt dt τv τv

βv dv Uv + U v τv dt



taking into account the water mass balance (13.3). The last quantity is the linear momentum of the material points changing phase from liquid water with velocity Ul to vapor with velocity Uv . The amount of mass changing phase and becoming vapor is

13.6 Entropy Balance

231



 βv d v βv ( ) + div Uv . dt τv τv

• The macroscopic angular momentum equations of motion

σv − σvT = 0, σl − σlT = 0, σa − σaT = 0, where σ T is the transposed matrix of matrix σ . They are satisfied if the constitutive laws give symmetric stresses. That is going to be the case in the sequel. • The microscopic motion equations of motion − Bvv + divHv = 0, in Ω , Hv · N = 0, on ∂ Ω , −Bll + divHl = 0, in Ω , Hl · N = 0, on ∂ Ω , −Bva = 0, in Ω .

(13.6)

These equations of motion are new. They describe the effects at the macroscopic level of the microscopic motions responsible for the liquid vapor phase changes, [107, 112].

13.6 Entropy Balance Because the specific volumes intervene, the results of Chap. 3 which assume the densities are constant are to be modified. The changes result from the derivatives of volume integrals, for instance the time derivative of the internal energy present in a volume Ω (t).

13.6.1 The Energy Balance The energy balance of the cloud system is: The actual work of the acceleration forces plus the variation of the internal energy is equal to the work of the exterior forces plus the quantity of heat provided to the system. The actual power of the acceleration forces is   v d βv

 βv Pacc (O) = ( Uv ) + ( Uv ) div Uv · Uv d Ω dt τv τv O    d l βl βl ( Ul ) + ( Ul ) div Ul · Ul d Ω + τl O dt τl

232

13 Clouds: Mixture of Air, Vapor and Liquid Water

  a d βv

 βv + ( Ua ) + ( Ua ) div Ua · Ua d Ω dt τa τa O    2    v dv βv U2v βv d βv Uv = d Ω (t) + ( ) + div Uv dΩ dτ τ 2 dt τ τ 2 Ov (τ ) v O v v    2   l Ul dl βl U2l βl d βl d Ω (t) + ( ) + div Ul dΩ + dτ τ 2 dt τ τ 2 Ov (τ ) l O l l   da βv U2a d Ω (t). + dτ Ov (τ ) τa 2 Denoting K =



β U2 dΩ , O(τ ) τ 2

the different kinetic energies, we get with the water mass balance (13.3) Pacc (O) =

dv Kv dl Kl da Ka + + + dτ dτ dτ

  v d βv O

(

dt τv

)+

βv div Uv τv



U2v U2l − 2 2

 dΩ .

Thus the energy balance is ∀t, ∀O, Ov (t) = Ol (t) = Oa (t) = O,  2    v d βv Uv U2l dv Kv dl Kl da Ka βv dΩ ( ) + div Uv − + + + dτ dτ dτ τv 2 2 O dt τv       dl da dv + ev d Ω + el d Ω + ea d Ω dτ dτ dτ Ov (τ ) Ol (τ ) Oa (τ ) = Pext (O) +



O

T Rext d Ω +



∂O

−T Q · NdΓ ,

(13.7)

where the e’s are the volumic internal energies, T Q is the heat flux vector, Q being the entropy flux vector and Rext T is the exterior volumic source of heat, Rext being the exterior volumic source of entropy. Relationship (13.7) means that the heat and work provided to the cloud modify the internal energy, the kinetic energy of each phase and adapt the kinetic energy of material changing phase to the kinetic energy of the new phase. Subtracting the principle of virtual power with the actual velocities relationship Pacc (O) = Pint (O) + Pext (O), from the energy balance (13.7), we get the classical relationship

13.6 Entropy Balance

233

∀t, ∀O, Ov (t) = Ol (t) = Oa (t) = O,       dl da dv ev (x, τ )d Ω + el (x, τ )d Ω + ea (x, τ )d Ω dτ dτ dτ Ov (τ ) Ol (τ ) Oa (τ ) = −Pint (O) +



O

T Rext d Ω +



∂O

−T Q · NdΓ .

It results the energy balance is 

  l   a  dv ev d el d ea + ev div Uv + + el div Ul + + ea div Ua dt dt dt

+ div T Q = F mec · δ E + Rext T, in Ω ,

(13.8)

−T Q · N = T π ext , in ∂ Ω , where T π ext is the exterior surfacic source of heat, π ext being the exterior surfacic source of entropy. We define vector F = (σv , σl , σa , Bvv , Bll , Bva , Hv , Hl , fvl , fva , fla , 0, 0, 0, 0, 0, −Q) = F mec + (16  0 , −Q).  times Vector F is equal to vector F mec where the last component of F mec which is equal to 0, is replaced by −Q. Relationship (13.8) gives 

  l   a  d el d ea dv ev + ev div Uv + + el div Ul + + ea div Ua dt dt dt

+T div Q = F · δ E + Rext T, in Ω , −T Q · N = T π ext , on ∂ Ω .

(13.9)

13.6.2 The Second Law It is ∀t, ∀O, Ov (t) = Ol (t) = Oa (t) = O,       dl da dv sv (x, τ )d Ω + sl (x, τ )d Ω + sa (x, τ )d Ω dτ dτ dτ Ov (τ ) Ol (τ ) Oa (τ ) ≥



R dΩ +



ext

O

∂O

−Q · Nd Γ ,

234

13 Clouds: Mixture of Air, Vapor and Liquid Water

and T > 0. It results 

(13.10)

  l   a  d v sv d sl d sa + sv div Uv + + sl div Ul + + ea div Ua + div Q dt dt dt

≥ Rext , in Ω .

(13.11)

13.6.3 The Free Energies Non dissipative interior forces are defined with a potential. We choose the volumic free energy.

13.6.3.1 Volumic Vapor Free Energy and Its Time Derivative It is

kv Ψ˜v (Ev ) = βvΨv (τv , T ) + I(βv) + (grad βv )2 + I0 (βv + βl − 1), 2 where Ψv (τv , T ) is given by physics, I is the indicator function of interval [0, 1] and I0 is the indicator function of the origin of R. Dependence on grad βv has been chosen quadratic for the sake of simplicity, with parameter kv non negative. This free energy implies that βv + βl − 1 = 0, βv ∈ [0, 1], βl ∈ [0, 1]. Water is either liquid or vapor (there is no ice). Moreover relationship βv + βl − 1 = 0 implies that there are no voids in the vapor liquid mixture and that no interpenetration occurs between liquid and vapor phases. In energy balance equations (13.9) and (13.14) below, the internal constraints on the state quantities are satisfied because the laws of mechanics apply for actual evolutions. Thus time derivatives dv ev /dt and dvΨ˜v /dt which intervene in (13.9) and (13.14) below are computed with the internal constraints on the state quantities which are satisfied. Thus in these relationships we have

∂Ψv dv τv d v βv dvΨ˜v = βv + Ψv dt ∂ τv dt dt +kv grad βv · grad

dv βv ∂Ψv dv T − kv grad βv ⊗ grad βv : grad Uv + βv , dt ∂ T dt

where we have used dv grad βv dv β v = grad − grad βv · grad Uv , dt dt

13.6 Entropy Balance

with and

235

(grad βv · gradUv ) j = (βv )i (Uv )i, j , (grad βv ⊗ grad βv )i j = (βv ),i (βv ), j ,

which is symmetric. The reactions to the internal constraints are defined by Breacvv ∈ ∂ I(βv ), Breacv ∈ ∂ I0 (βv + βl − 1). They satisfy, following Theorem 3.2 (Breacvv + Breacv )

d v βv dv β l + Breacv ≥ 0. dt dt

Remark 13.4. The reactions depend on the values of βv and of βv + βl − 1 but they depend also on x and t: they are Breacvv (βv , x,t) and Breacv (βv + βl − 1, x,t). In the sequel we write Breacvv and Breacv instead of Breacvv (βv , x,t) and of Breacv (βv + βl − 1, x,t). Let us recall that if the evolution is smooth (see definition in Remark 3.9), the inequality is an equality and the reactions are workless. For instance, this is the case, if either functions t → β are differentiable or reactions t → Breacvv and t → Breacv are continuous.

13.6.3.2 Volumic Liquid Water Free Energy and Its Time Derivative It is

k Ψ˜l (El ) = βlΨl (τl , T ) + I(βl ) + l (grad βl )2 + I0(βv + βl − 1), 2 where Ψl (τl , T ) is given by physics.

Remark 13.5. The dependence of Ψ˜v and Ψ˜l on grad βv and grad βl has been chosen quadratic for the sake of simplicity. In a less schematic theory, we may choose a non quadratic dependence with kv = kv (τv , T ) and kl = kl (τl , T ). Moreover we may assume kv = kl . This assumption is made in the sequel, see formula (13.28). In an actual evolution, the time derivative of the liquid water free energy satisfies

∂Ψl dl τl dl β l dlΨ˜l = βl + Ψl dt ∂ τl dt dt +kl grad βl · grad

dl βl ∂Ψl dl T − kl grad βl ⊗ grad βl : gradUl + βl . dt ∂ T dt

236

13 Clouds: Mixture of Air, Vapor and Liquid Water

The reactions to the internal constraints on the state quantities are defined by Breacll ∈ ∂ I(βv ), Breacl ∈ ∂ I0 (βv + βl − 1). They satisfy, following Theorem 3.2 (Breacll + Breacl )

d l βl dl β v + Breacl ≥ 0, dt dt

the inequality being an equality in case the evolution is smooth (see Remark 3.9).

13.6.3.3 Volumic Air Free Energy It is

Ψ˜a (El ) = βvΨa (τa , T ) + I(βv ),

where Ψa (τa , T ) is given by physics, for instance

Ψa (τa , T ) = −rT

ln τa , τa

(13.12)

where r is a constant. This free energy is the ideal gas free energy. We do not add the indicator function I0 (βv + βl − 1) because air does not change phase, (βv being the volume fraction occupied by the air). In an actual evolution, the time derivative of the air free energy satisfies daΨ˜a ∂Ψa da τa d v βv ∂Ψa da T = βv + Ψa + βv . dt ∂ τa dt dt ∂ T dt The reaction to the internal constraint is defined by Breacva ∈ ∂ I(βv ), which satisfies, following Theorem 3.2 Breacva

dv β v ≥ 0, dt

the inequality being an equality in case the evolution is smooth (see Remark 3.9). Remark 13.6. The three specific volumes τ are positive quantities. This physical property is to be taken into account by each free energy Ψ by adding to Ψ the indicator function I+ (τ ) of R+ , Ψ = Ψ + I+ (τ ). For the sake of simplicity, we do not mention this property in the sequel. The free energies we are going to choose will

13.6 Entropy Balance

237

satisfy lim Ψ (τ ) = ∞. Thus if specific volume is non negative at time 0, it remains τ →0

non negative. Let us note that the specific free energies are Ψˆv = τvΨv and Ψˆl = τlΨl .

13.6.3.4 The Non Dissipative Interior Forces and the Non Dissipative Reactions Forces We define the non dissipative interior forces vector F nd with the derivatives of free energies Ψ˜v , Ψ˜l and Ψ˜a . It is  F nd = σvnd = −kv grad βv ⊗ grad βv + βvΨv 1,

σlnd = −kl grad βl ⊗ grad βl + βlΨl 1, σand = βvΨa 1, nd Bnd vv = Ψv , Bll = Ψl ,

Bnd va = Ψa , nd Hnd v = kv grad βv , Hl = kl grad βl , nd fnd vl = 0, fva = 0,

fnd la = 0,

∂Ψv nd ∂Ψl nd ∂Ψa , Z = βl , Z = βv , ∂ τv l ∂ τl a ∂ τa nd = 0, Bˆ nd lv = 0, −Q = 0 ,

Zvnd = βv Bˆ nd vl

where 1 is the identity matrix. We define the non dissipative reaction forces vector F reac with the reactions to the internal constraints taken into account by the free energies. It is F reac = {σvreac = 0, σlreac = 0, σareac = 0, reac Breac = Breacll + Breacl , vv = Breacvv + Breacv , Bll

Breac va = Breacva , = 0, Hreac = 0, Hreac v l = 0, freac freac va = 0, vl freac la = 0, Zvreac = 0, Zlreac = 0, Zareac = 0, Bˆ reac = Breacv , Bˆ reac vl lv = Breacl , −Qreac = 0} ,

238

13 Clouds: Mixture of Air, Vapor and Liquid Water

Let us recall that the entropies are (see Sect. 3.2.3) sv = −βv

∂Ψv ∂Ψl ∂Ψa , sl = −βl , sa = −βv . ∂T ∂T ∂T

With the computations of the time derivatives of the free energies and the definition of F nd , we get d sT dl T da T + sl + sa dt dt dt v dlΨ˜l daΨ˜a d Ψ˜v + + Ψ˜v div Uv + + Ψ˜l div Ul + + Ψ˜a div Ua dt dt dt sv

= F nd · δ E. We have also

(13.13) F reac · δ E ≥ 0.

Let us recall that assuming smooth evolution, i.e., the non dissipative reactions are workless, the inequality becomes an equality (see Remark 3.9). This is why vector F reac is called non dissipative reaction vector.

13.6.4 The Entropy Balance Following the definition of the free energy given in Sect. 3.2.3, we use relationship e = Ψ˜ + T s, and get from energy balance (13.9)  T

d v sv dl sl da sa + sv div Uv + + sl div Ul + + sa div Ua + divQ dt dt dt



dsT dl T da T + sl + sa dt dt dt v dlΨ˜l daΨ˜a d Ψ˜v + + Ψ˜v div Uv + + Ψ˜l div Ul + + Ψ˜a div Ua dt dt dt

+sv

= F · δ E + Rext T, in Ω ,

(13.14)

By relationship (13.13), we get 

d v sv dl sl da sa + sv div Uv + + sl div Ul + + sa div Ua + divQ dt dt dt 

= F − F nd · δ E + Rext T, in Ω ,



T

(13.15)

13.6 Entropy Balance

239

The interior forces and entropy flux vector F is split between a dissipative part F d with index d , a non dissipative part F nd with index nd and a reaction part F reac with index reac F = F d + F nd + F reac , with F = (σv , σl , σa , Bvv , Bll , Bva , Hv , Hl , fvl , fva , fla , 0, 0, 0, 0, 0, −Q),

F d = (σvd , σld , σad , Bdvv , Bdll , Bdva , Hdv , Hdl , fdvl , fdva , fdla , Zvd , Zld , Zad , Bˆ dvl , Bˆ dlv , −Qd ), nd nd nd nd F nd = (σvnd , σlnd , σand , Bnd vv , Bll , Bva , Hv , Hl , nd nd nd nd nd ˆ nd ˆ nd nd fnd vl , fva , fla , Zv , Zl , Za , Bvl , Blv , −Q ),

and reac reac reac reac F reac = (σvreac , σlreac , σareac , Breac , vv , Bll , Bva , Hv , Hl reac reac reac reac reac ˆ reac ˆ reac , Za , Bvl , Blv , −Qreac ). freac vl , fva , fla , Zv , Zl

It results from energy balance (13.15) and relationship (13.10) 

  l   a  dv sv d sl d sa + sv div Uv + + sl div Ul + + sa div Ua + divQ dt dt dt = D + D reac + Rext , in Ω , −Q · N = π ext , on ∂ Ω ,

(13.16)

where

 1 δ E · Fd , T the volume interior rate of entropy production is as usual the product of δ E by the dissipative forces F d and D=

D reac =

1 (δ E · F reac ) ≥ 0, T

(13.17)

is the non dissipative reaction forces entropy production which is null when the evolution is smooth. We have already assumed this property and have the entropy balance

240

13 Clouds: Mixture of Air, Vapor and Liquid Water



  l   a  dv sv d sl d sa + sv div Uv + + sl div Ul + + sa div Ua + divQ dt dt dt

= D + Rext , in Ω , −Q · N = π ext , on ∂ Ω ,

(13.18)

which is equivalent to the energy balance.

13.7 The Second Law: An Equivalent Formulation By comparing relationships (13.11) and (13.18), we get an equivalent formulation of the second law T > 0, D ≥0.

(13.19)

It is satisfied if T > 0 (which we assume or which may be ensured by having the free energies to depend on ln T ) and if constitutive law for F d is defined by a pseudopotential of dissipation Φ (E, δ E) F d ∈ ∂ Φ (E, δ E),

(13.20)

where the subdifferential set is computed with respect to δ E. The proof is given in Theorem 3.1. Remark 13.7. In case the evolution is not smooth, second law is T > 0, D + D reac ≥0. Let us note that constitutive law (13.20) and relationship (13.17) ensure the second law is satisfied in this situation.

13.8 Pseudopotential of Dissipation The pseudopotential of dissipation describes the dissipative properties which depend on the velocities and on the thermal heterogeneity, δ E. Among these properties, there are the internal constraints on the velocities.

13.8 Pseudopotential of Dissipation

241

13.8.1 Internal Constraints on the Velocities We have internal constraints on velocities   ∂ βv ∂ βv dl βv = + Ul · grad βv = + Uv · grad βv + (Ul − Uv ) · grad βv dt ∂t ∂t =

d v βv + (Ul − Uv ) · grad βv , dt

and

dv β l dl β l = + (Uv − Ul ) · grad βl . dt dt These two internal constraints with βv + βl = 1 are equivalent to 1 dl βv dv βl + = (Uv − Ul ) · grad(βl − βv ), dt dt 2 βv + βl = 1.

(13.21)

We have also dv β v da βv = + (Ua − Uv ) · grad βv . dt dt

(13.22)

Constraints (13.21) and (13.22) involve velocities. Thus they are taken into account by the pseudopotential of dissipation. In the same way, mass balances (13.3) and (13.4) are constraints on the velocities which are taken into account by the pseudopotential of dissipation.

13.8.2 The Pseudopotential of Dissipation We choose 

     dl β l dv βv βv βl + + div Uv + div Ul Φ (E, δ E) = I0 dt τv dt τl τv τl  a   βv βv d +I0 + div Ua dt τa τa   l d βv d v β l 1 +I0 + − (Uv − Ul ) · grad(βl − βv ) dt dt 2  a  v d βv d β v +I0 − − (Ua − Uv ) · grad βv dt dt

242

13 Clouds: Mixture of Air, Vapor and Liquid Water

βv βl kvl βv kva βv βl kla (Uv − Ul )2 + (Uv − Ua )2 + (Ul − Ua )2 2 2 2 βl  + λl (div Ul )2 + 2μl D(Ul ) : D(Ul ) 2 βv  + λv (div Uv )2 + 2μv D(Uv ) : D(Uv ) 2 βv  + λa (div Ua )2 + 2μa D(Ua ) : D(Ua ) 2 λ + (grad T )2 . 2T +

We assume air, vapor and liquid water to be viscous. The λi and μi are the viscosity parameters. The quadratic terms as (Uv − Ul )2 account for frictions between the phases, the intensity of which is proportional to the velocity difference (Uv − Ul ). We have assume thermal conductivity λ to be the same in any phase, it is possible to have λ = λ (βl , βv ) depending on the volume fractions.

13.9 Constitutive Laws As usual and already done, the internal forces F are split between non dissipative forces F nd , non dissipative reaction forces F reac and dissipative forces F d . Forces F nd and F reac have already been defined.

13.9.1 The Dissipative Internal Forces They are defined by relationship (13.20) F d ∈ ∂ Φ (E, δ E). We define the specific pressures pw , pa and reaction works p and q  −pw ∈ ∂ I0  −pa ∈ ∂ I0

dv dt

da dt

 

βv τv βv τa

 +  +

dl dt



βl τl

 +

 βv div Ua , τa

 βv βl div Uv + div Ul , τv τl

13.9 Constitutive Laws

243

 d l β v d v βl 1 + − (Uv − Ul ) · grad(βl − βv ) , −p ∈ ∂ I0 dt dt 2  a  v d β v d βv −q ∈ ∂ I0 − − (Ua − Uv ) · grad βv . dt dt 

It results vector F d is  βv F d = σvd = −pw 1 + βv {λv div Uv 1 + 2μvD(Uv )} , τv

σld = −pw

βl 1 + βl {λl div Ul 1 + 2μl D(Ul )} , τl

σad = −pa

βv 1 + βv {λa div Ua 1 + 2μaD(Ua )} , τa

Bdvv = −pw

1 1 + q, Bdll = −pw , τv τl

Bdva = −pa

1 − q, τa

Hdv = 0, Hnd l = 0, fdvl = βl βv kvl (Uv − Ul ) −

p grad(βl − βv ), fdva = βv kva (Uv − Ua ) − q grad βv , 2

fdla = βl βv kla (Ul − Ua ), Zvd = pw

βv d βl βv , Zl = pw 2 , Zad = pa 2 , 2 τv τa τl

Bˆ dvl = −p, Bˆ dlv = −p,  λ −Qd = grad T . T

13.9.2 The Constitutive Laws They are given by vector F F = F d + F nd + F reac ,

(13.23)

244

13 Clouds: Mixture of Air, Vapor and Liquid Water

the components of which are

σv = −pw

βv 1 + βvΨv (τv , T )1 + βv {λv div Uv 1 + 2μvD(Uv )} τv

−kv grad βv ⊗ grad βv ,

σl = −pw

βl 1 + βlΨl (τl , T )1 + βl {λl div Ul 1 + 2μl D(Ul )} − kl grad βl ⊗ grad βl , τl

βv 1 + βvΨa (τa , T )1 + βv {λa div Ua 1 + 2μaD(Ua )} , τa 1 = −pw + q + Ψv(τv , T ) + Breacvv + Breacv, τv

σa = −pa Bvv

Bll = −pw

1 + Ψl (τl , T ) + Breacll + Breacl , τl

Bva = −pa

1 − q + Ψa(τa , T ) + Breacva , τa

Hv = kv grad βv , Hl = kl grad βl , p fvl = βl βv kvl (Uv − Ul ) − grad(βl − βv ), fva = βv kva (Uv − Ua ) − q grad βv , 2 fla = βl βv kla (Ul − Ua ), 0 = pw

βv ∂Ψv βl ∂Ψl βv ∂Ψa + βv , 0 = pw 2 + βl , 0 = p a 2 + βv , τv2 ∂ τv ∂ τ τ ∂ τa τl l a

0 = −p + Breacv, 0 = −p + Breacl , −Q =

λ grad T. T

(13.24) (13.25) (13.26)

13.9.3 Consequences of the Constitutive Laws We have

βv ∈ [0, 1], βl ∈ [0, 1] , βv + βl = 1,

because there exist reaction forces which imply that the different subdifferential sets are not empty, [112], see also Theorem A.2 of Appendix A. The constitutive laws imply also that mass balances (13.3), (13.4) and internal constraints (13.21), (13.22) on the velocities are satisfied. This is also a consequence of the existence of the dissipative reactions forces, the specific pressures, pw and pa , and reaction works, p and q.

13.10 The Equations of the Predictive Theory

245

13.10 The Equations of the Predictive Theory They result from the equations of motion, (13.5) and (13.6), the entropy balance, (13.18) and the constitutive laws, (13.23). Note again that the constitutive laws imply mass balances (13.3), (13.4) and internal constraints (13.21), (13.22) on the velocities are satisfied. They imply also the constraints on the volume fractions are satisfied. With relationships (13.25), it is possible to eliminate Breacv and Breacl from the equations for the microscopic motion   1 − Ψv (τv , T ) − pw − Breacvv − p − q + kvΔ βv = 0, τv   1 − Breacll − p + kl Δ βl = 0, − Ψl (τl , T ) − pw τl   1 − Breacva + q = 0. − Ψa (τa , T ) − pa τa By adding and subtracting the two first equations and eliminating q, we get       1 1 1 − Ψv (τv , T ) − pw + Ψl (τl , T ) − pw − Ψa (τa , T ) − pa τa τl τv −Breacva − Breacvv + Breacll + kv Δ βv − kl Δ βl = 0,       1 1 1 − Ψl (τl , T ) − pw − Ψa (τa , T ) − pa − Ψv (τv , T ) − pw τl τv τa −Breacvv − Breacll − 2p + kvΔ βv + kl Δ βl = 0,   1 − Breacva + q = 0. − Ψa (τa , T ) − pa τa Using relationship βv + βl = 1, we get       1 1 1 − Ψv (τv , T ) − pw + Ψl (τl , T ) − pw − Ψa (τa , T ) − pa τa τv τl −Breacvv + Breacll + (kv + kl )Δ βv = 0,       1 1 1 − Ψv (τv , T ) − pw − Ψl (τl , T ) − pw − Ψa (τa , T ) − pa τl τv τa −Breacvv − Breacll − 2p + (kv − kl )Δ βv = 0,   1 − Breacva + q = 0. − Ψa (τa , T ) − pa τa

246

13 Clouds: Mixture of Air, Vapor and Liquid Water

For the sake of simplicity we choose kv = kl = kβ , (see Remark 13.5). By noting that Breacvv − Breacll ∈ ∂ I(βv ) − ∂ I(1 − βv) = ∂ I(βv ), we get       1 1 1 Ψa (τa , T ) − pa + Ψv (τv , T ) − pw − Ψl (τl , T ) − pw τa τv τl (13.27) +Breac − 2kβ Δ βv = 0,       1 1 1 − Ψl (τl , T ) − pw − Ψa (τa , T ) − pa − Ψv (τv , T ) − pw τv τl τa −Breacvv − Breacll = 2p,   1 − Ψa (τa , T ) − pa − Breacva + q = 0. τa

(13.28) (13.29)

Breac ∈ ∂ I(βv ), Breacvv ∈ ∂ I(βv ), Breacll ∈ ∂ I(1 − βv), Breacva ∈ ∂ I(βv ). The other equations are  βv + div Uv τv    1 = − div(k grad βv ⊗ grad βv ) + grad βv Ψv − pw τv

βv dv Uv + Uv τv dt



dv dt



βv τv



+βv {(λv + μv ) grad div Uv + μv Δ Uv } p −βl βv kvl (Uv − Ul ) − grad(1 − 2βv) − βv kva (Uv − Ua ) + q grad βv + fext v , 2  l   1 − βv d 1 − βv 1 − β v dl + Ul + Ul div Ul τl dt dt τl τl    1 = − div(k grad βv ⊗ grad βv ) + grad (1 − βv) Ψl − pw τl +(1 − βv) {(λl + μl ) grad div Ul + μl Δ Ul } p +βv βl kvl (Uv − Ul ) + grad(1 − 2βv) − βv βl kla (Ul − Ua ) + fext l , 2    βv da 1 Ua = grad βv Ψa − pa τa dt τa

(13.30)

+βv {(λa + μa ) grad div Ua + μaΔ Ua } +βv βl kla (Ul − Ua ) + βv kva (Uv − Ua ) − q grad βv + fext a ,

(13.31)

13.10 The Equations of the Predictive Theory

 0 = βv  = βv

∂Ψv 1 + pw 2 ∂ τv τv ∂Ψa 1 + pa 2 ∂ τa τa

 

247



∂Ψl 1 + pw 2 = (1 − βv) ∂ τl τl



,

(13.32)

dv sv dl sl da sa + sv div Uv + + sl div Ul + + sa div Ua dt dt dt −λ Δ ln T = D + Rext ,     dv βv 1 − βv βv dl 1 − βv div Ul = 0, + + div Uv + dt τv dt τl τv τl   βv da β v + div Ua = 0. dt τa τa

(13.33) (13.34) (13.35)

External actions fvext , flext , faext and Rext are given, for instance fvext =

1 − βv βv βv g, flext = g, faext = g, Rext = 0, τv τl τv

where g is the gravity acceleration vector. There are 18 equations, (13.27)–(13.35), for 9 velocities U, 3 specific volumes τ , 1 volume fraction βv , 1 temperature T , 2 pressures pw , pa , and 2 reaction works p, q, unknowns, i.e., for 18 unknowns. Boundary and initial conditions are to be added. The mechanical novelty is the local interaction in the liquid water-vapor phase change: the volume fractions at a material point influence the neighbourhood volume fractions. This interaction introduces the gradients of the βi ’s which result in the laplacian in (13.27) for βv . Remark 13.8. Equations (13.28)–(13.29) give reaction works p and q with value: • If 0 < βv < 1,   1 2p = − Ψv (τv , T ) − pw τv     1 1 − Ψl (τl , T ) − pw − Ψa (τa , T ) − pa , τl τa   1 . q = Ψa (τa , T ) − pa τa

(13.36) (13.37)

In the liquid vapor air mixture, it is possible that there is no vapor by having τv = ∞. Note that we have   1 lim Ψv (τv , T ) − pw = 0, τv →∞ τv

248

13 Clouds: Mixture of Air, Vapor and Liquid Water

because vapor with large τv behaves like an ideal gas. Equation (13.27) becomes     1 1 Ψa (τa , T ) − pa − Ψl (τl , T ) − pw + Breac − 2kβ Δ βv = 0. τa τl It is also possible that there is no air by having τa = ∞ with   1 lim Ψa (τa , T ) − pa = 0. τa →∞ τa Equation (13.27) becomes     1 1 Ψv (τv , T ) − pw − Ψl (τl , T ) − pw + Breac − 2kβ Δ βv = 0. τv τl • If βv = 1, there is only air and vapor. The reaction works are   1 . p ∈ R, q ≥ Ψa (τa , T ) − pa τa It is possible that the air vapor mixture contains either only vapor with τa = ∞ or only air with τv = ∞ (see Sect. 13.11.4 down below). • If βv = 0, there is no air which can only appear by diffusion with τa = ∞, thus we have   1 = 0, p ∈ R, q ≤ lim Ψa (τa , T ) − pa τa →∞ τa because air with large τa behaves like an ideal gas. In case βv is 0 (there is only liquid water), or in case βv is 1 (there is only vapor and air), reaction work p does not intervene in the equations (because grad (1 − 2βv) = 0). Reaction q does not intervene either because grad βv = 0. Remark 13.9. Equation (13.32) may by replaced by their sum and their difference, giving − pw = βv τv2

∂Ψv ∂Ψl + (1 − βv)τl2 , ∂ τv ∂ τl

−pw (2βv − 1) = βv τv2

∂Ψv ∂Ψl − (1 − βv)τl2 . ∂ τv ∂ τl

(13.38)

First relationship gives water specific pressure pw which can be eliminated and reducing the number of unknowns to 17.

13.10 The Equations of the Predictive Theory

249

13.10.1 A Potential and the Free Enthalpy Free energies Ψ are assumed to be convex functions of densities 1/τ . Because quantity p/τ − Ψ is involved in numerous formulas, it may useful to introduce new potentials defined by  1 = sup pv − Ψv (τ , T ) , − Gv (pv , T ) = τ τ   1 −Gl (pv , T ) = Ψl∗ (pv , T ) = sup pl − Ψl (τ , T ) , τ τ

Ψv∗ (pv , T )



(13.39)

(see Sect. A.5 of Appendix A), and have constitutive laws

∂Ψv∗ ∂ Gv 1 (pv , T ) = − = (pv , T ), τv ∂ (pv ) ∂ pv 1 ∂ Gl =− (pl , T ), τl ∂ pl together with pv = giving

∂Ψv ∂Ψv ∂Ψl (τv , T ), pl = −τl2 (τl , T ), (τv , T ) = −τv2 ∂ (1/τv ) ∂ τv ∂ τl pw = βv pv + (1 − βv)pl .

Remark 13.10. The schematic free energies chosen below are convex functions of the densities 1/τ . Potential Ga with free energy given by (13.12) is Ga (pa , T ) = −kT exp( giving

pa − kT ), kT

pa − kT 1 ), pa = kT − kT ln τa . = exp( τa kT

Let us note that if 1/τ → Ψ (τ , T ) is convex and if τ ≥ 0 (see Remark 13.6), specific free energy τ → Ψˆ (τ , T ) = τΨ (τ , T ) is also convex Theorem 13.1. If τ ≥ 0 convexity of function 1/τ → Ψ (τ , T ) is equivalent to convexity of function τ → Ψˆ (τ , T ) = τΨ (τ , T ). Proof. It results from relationship

∂ 2Ψ ∂ 2Ψ ∂Ψ ∂ 2Ψˆ = τ 3 (2 + τ 2 ) = τ3 2 . 2 ∂ (1/τ ) ∂τ ∂τ ∂τ

250

13 Clouds: Mixture of Air, Vapor and Liquid Water

Let us recall that condition τ ≥ 0 has been included in the definition of the free energies Ψ , see Remark 13.6. 

ˆ T ) is defined by Within this convexity assumption, free enthalpy G(P,   ˆ T ) = sup −Px − Ψˆ (x, T ) , −G(P, x

with ˆ T ) = Pτ + Ψˆ (τ , T ), G(P, P=−

∂ Ψˆ ∂ Gˆ (τ , T ), τ = (P, T ). ∂τ ∂P

Let us note that

Ψ (τ , T ) − p

1 1 ∂Ψ ∂G = G(p, T ) ⇔ p = −τ 2 (τ , T ) ⇔ = − (p, T ), τ ∂τ τ ∂p

ˆ T) ⇔ P = − Pτ + Ψˆ (τ , T ) = G(P,

∂ Ψˆ ∂ Gˆ (P, T ). (τ , T ) ⇔ τ = ∂τ ∂P

(13.40)

These relationships imply

∂ (τΨ ) 1 (τ , T ) = p − Ψ (τ , T ) = −G(p, T ), ∂τ τ ∂Ψ ˆ T ). (τ , T ) = Ψˆ (τ , T ) + τ P = G(P, p = −τ 2 ∂τ

P=−

(13.41)

Remark 13.11. We have

∂p ∂ =− ∂τ ∂τ

  ∂ 2Ψ ∂ 2Ψˆ ∂Ψ 2 ∂Ψ τ + τ 2 ) = −τ 2 , = −τ (2 ∂τ ∂τ ∂τ ∂τ

∂P ∂ 2Ψˆ =− 2. ∂τ ∂τ It results that pressures τ → p(τ , T ) and τ → P(τ , T ) are decreasing functions of τ . Remark 13.12. Specific free energy Ψˆ of a material is defined up to a linear function of temperature AT + B. It results volume free energy Ψ = Ψˆ /τ is defined up to function (AT + B)/τ . These functions are the same for the free energies of all the phases, Ψˆl , Ψl and Ψˆv , Ψv . Remark 13.13. The air free enthalpy is rT Gˆ a (Pa , T ) = rT (1 − ln ), Pa with specific free energy Ψˆ (τa , T ) = −rT ln τa .

13.11 Air, Liquid Water and Vapor in an Homogeneous Cloud

251

13.11 Air, Liquid Water and Vapor in an Homogeneous Cloud In an homogeneous cloud, (13.27) governing the evolution of the mixture is       1 1 1 Ψa (τa , T ) − pa + Ψv (τv , T ) − pw ) − Ψl (τl , T ) − pw + Breac = 0, τa τv τl Breac ∈ ∂ I(βv ).

13.11.1 The Air, Liquid Water, Vapor Mixture In this case, we have 0 < βv < 1 and       1 1 1 Ψa (τa , T ) − pa + Ψv (τv , T ) − pw ) − Ψl (τl , T ) − pw = 0, (13.42) τa τv τl and, from relationships (13.32) 0=

∂Ψv 1 ∂Ψl 1 + pw 2 , 0 = + pw 2 . ∂ τv τv ∂ τl τl

By eliminating pw and pa we get

∂ (τaΨa ) ∂ (τvΨv ) ∂ (τlΨl ) + − = 0, ∂ τa ∂ τv ∂ τl −τv2

∂Ψv ∂Ψl = −τl2 . ∂ τv ∂ τl

(13.43) (13.44)

There are two relationships and four parameters (τa , τv , τl , T ). The two relationships (13.43) and (13.44) give

τv = τˆv (T, τa ), τl = τˆl (T, τa ), and the specific pressure pˆ w (T, τa ) = −τˆv2 (T, τa )

∂Ψv (τˆv (T, τa ), T ) ∂ τv

= −τˆl2 (T, τa )

∂Ψl (τˆl (T, τa ), T ). ∂ τl

(13.45)

The curves T → τˆv (T, τa ), T → τˆl (T, τa ) for τa fixed, are drawn in the thermodynamical diagrams. An example is given in Fig. 12.3 for the case there is no air

252

13 Clouds: Mixture of Air, Vapor and Liquid Water

(τa = ∞, {−Ψa (τa , T ) + pa /τa } = 0) where function T = τˆv−1 (τ ) is the green line and function T = τˆl−1 (τ ) is the blue line.

13.11.2 The Cloud Pressure When the cloud is homogeneous, pressure Pcloud in the cloud results from the stresses σv = −βv (pw 1/τv − Ψv ) 1, σl = −βl (pw /τl − Ψl ) 1 and σa = −βv (pa /τa − Ψa) 1. It is 

Pcloud

     1 1 1 = βv pw − Ψv + βl pw − Ψl + βv pa − Ψa τv τl τa       ∂Ψv ∂Ψl ∂Ψa + Ψv − βl τl + Ψl − βv τa + Ψa = −βv τv ∂ τv ∂ τl ∂ τa = − βv

∂ Ψˆv ∂ Ψˆl ∂ Ψˆa − βl − βv , ∂ τv ∂ τl ∂ τa

(13.46)

where Ψˆ = τΨ is the specific free energy and relationship (13.45) has been used. We define     1 ∂ Ψˆv 1 ∂ Ψˆl , Pl = pw − Ψl = − , Pv = pw − Ψv = − τv ∂ τv τl ∂ τl   1 ∂ Ψˆa Pa = pa − Ψa = − , τa ∂ τa the pressures in each material. We get with relationship (13.42): • In case the mixture contains liquid, vapor and air  Pcloud =

     1 1 1 + p = p pw − Ψv − Ψa − Ψl a w τv τa τl

= Pv + Pa = Pl . • In case the mixture contains only liquid  Pcloud =

1 pw − Ψl τl

 = Pl .

• In case the mixture contains vapor and air  Pcloud =

pw

   1 1 − Ψv + pa − Ψa = Pv + Pa . τv τa

13.11 Air, Liquid Water and Vapor in an Homogeneous Cloud

If there is no air,

 Pcloud =

pw

1 − Ψv τv

253

 = Pv .

The vapor liquid mixture specific pressure pw given by formulas (13.38), depend on (T, τa ). Pressure Pcloud depends also on (T, τa ) because τv = τˆv (T, τa ) and τl = τˆl (T, τa ) depend also on (T, τa ). Thus pw and Pcloud are constant when (T, τa ) are constant. Pressure Pcloud is shown on the Clapeyron water diagram of Fig. 12.3 in case there is no air (τa = ∞, (pa 1/τa − Ψa ) = 0). Let us note, that pressure Pcloud and pw are also constant on a free boundary where phase change occurs as seen in Chap. 12. Remark 13.14. Relationships (13.43) and (13.44) become Pl − Pv − Pa = 0,

Ψˆv + Pv τv = Ψˆl + Pl τl = pw .

(13.47)

These formulas are similar to formulas (12.2) and (12.4) of the surfacic phase change. Relationships (13.41) and (13.47) give Pv = Pˆv (T, τa ) = −Gv (pw (T, τa ), T ), Pl = Pˆl (T, τa ) = −Gl (pw (T, τa ), T ).

(13.48)

pw (T, τa ) = Gˆ v (Pˆv (T, τa ), T ) = Gˆ l (Pˆl (T, τa ), T ).

(13.49)

and

The free enthalpies are equal in the mixture. The specific pressures pl and pv are equal to pw whereas pressures Pv and Pl are not equal because of the air. They are equal when there is no air.

13.11.3 The Liquid Water In this case, we have βv = 0 and       1 1 1 Ψa (τa , T ) − pa + Ψv (τv , T ) − pw ) − Ψl (τl , T ) − pw > 0. τa τv τl (13.50) If some vapor is to appear at temperature T it may have any specific volume τv . If some air is to appear, it appears by diffusion. Thus when appearing its density is 0, i.e., its specific volume τa is infinite. Thus condition (13.50) has to be satisfied for any specific volume τv and τa = ∞. It results

254

13 Clouds: Mixture of Air, Vapor and Liquid Water

    1 1 inf Ψv (τv , T ) − pl (τl , T ) > Ψl (τl , T ) − pl (τl , T ) τv τv τl   1 . − lim Ψa (τa , T ) − pa τa →0 τa with pl (τl , T ) = −τl2

∂Ψl (τl , T ). ∂ τl

Note that air being an ideal gas for large τa , we have   1 = 0, Ψa (τa , T ) − pa τa →0 τa lim

and the air does not intervene in the liquid vapor phase change. We have due to definition (13.39) of potentials G   1 inf Ψv (τv , T ) − pl (τl , T ) = Gv (pl (τl , T ), T ), τv τv   1 = Gl (pl (τl , T ), T ). Ψl (τl , T ) − pl (τl , T ) τl Thus condition (13.50) is Gv (pl (τl , T ), T ) > Gl (pl (τl , T ), T ). Function p → Gv (p, T ) − Gl (p, T ) is an increasing function because 1 1 ∂ Gv ∂ Gl (p, T ) − (p, T ) = − + > 0. ∂p ∂p τv (p, T ) τl (p, T ) Note that we have Gv ( pˆw (T ), T ) − Gl ( pˆw (T ), T ) = 0, where specific pressure pˆw is the liquid vapor mixture specific pressure pˆw (T ) = pˆw (T, τa = ∞). Thus condition (13.50) is equivalent to Gv (pl (τl , T ), T ) − Gl (pl (τl , T ), T ) > Gv ( pˆw (T ), T ) − Gl ( pˆw (T ), T ) = 0, or to pl (τl , T ) > pˆw (T ).

(13.51)

Because pressure pl is a decreasing function of specific volume (see Remark 13.11), condition (13.50) is equivalent to

13.11 Air, Liquid Water and Vapor in an Homogeneous Cloud

255

τl < τˆl (T, τa = ∞) ⇔ pl (τl , T ) > pˆw (T ). Due to relationships (13.41) and (13.47), relationship (13.51) is equivalent to Gˆ l (Pcloud , T ) > Gˆ l (Pˆl (T ), T ), where Pcloud = Pl and Pˆl (T ) is the liquid pressure in the vapor, liquid water mixture at temperature T . Because free enthalpy Gˆ is an increasing function of P ˆ ∂ P = τ > 0), the previous inequality is equivalent to (relationship (13.40) gives ∂ G/ Pcloud > Pˆl (T ), see diagram of Fig. 12.1.

13.11.4 The Air, Vapor Mixture In this case, we have βv = 1 and       1 1 1 + Ψv (τv , T ) − pw ) − Ψl (τl , T ) − pw < 0. (13.52) Ψa (τa , T ) − pa τa τv τl This condition is       1 1 1 Ψa (τa , T ) − pa + Ψv (τv , T ) − pv (τv , T ) < inf Ψl (τl , T ) − pv (τv , T ) , τl τa τv τl or

  1 Gv (pv (τv , T ), T ) + Ψa (τa , T ) − pa < Gl (pv (τv , T ), T ). τa

With relationship   1 Gv ( pˆw (T, τa ), T ) + Ψa (τa , T ) − pa = Gl ( pˆw (T, τa ), T ), τa and relationships (13.40), (13.47), condition (13.52) is equivalent to pv (τv , T ) < pˆ w (T, τa ) ⇔ τv > τˆv (T, τa ) ⇔ Pv < Pˆv (T, τa ) ⇔ Pcloud < Pˆv (T, τa ) + Pa (T, τa ) ⇔ Gˆ v (Pv , T ) < Gˆ v (Pˆv (T, τa ), T ),

256

13 Clouds: Mixture of Air, Vapor and Liquid Water

where specific pressure pˆw (T, τa ) is the air liquid vapor mixture specific pressure and Pˆv (T, τa ) is the vapor pressure in the air vapor liquid water mixture at temperature T with air specific volume τa . It is possible that there is no air in the air vapor mixture with τa = ∞, Ψa (τa , T ) − pa /τa = 0 and Pa (T, τa ) = 0, see diagram of Fig. 12.1. The effect of the air is to favor the presence of vapor because air pressure Pa (T, τa ) is large when there is much air, i.e., τa is small. In case there is only air and no vapor, τv = ∞ and Gv (pv (τv = ∞), T ) = 0, with ideal gas specific pressure lim pv (τv ) = −∞ and Pv = 0. The previous relationships τv =∞

are satisfied. In this situation vapor can only appear by diffusion because there is neither liquid nor vapor. Of course, it is impossible to have both air and vapor missing in the mixture (we have assumed that it is impossible to have voids, i.e., βv + βl = 1).

13.12 A Schematic Example in Case There Is No Air For T ≤ Tc the two phases may coexist. Temperature Tc is the critical temperature. We choose the following schematic free energies

Ψl (τl , T ) = −kT

ln τl τc − τl +d (T − Tc ), τl τl

Ψv (τv , T ) = −kT

ln τv , τv

where k > 0, τc > 0 and d > 0 are physical constants. In diagrams 13.1 and 13.2, their values are Tc = 650 K, τc = 3 × 10−3 m3 / kg, k = 1.001 × 105 J/(K m3 ) and d = 5.05 × 107 J/(K m3 ). Remark 13.15. Note that the free energies are convex functions of 1/τ , giving potentials G concave functions of pressures. The specific free energies Ψˆ = τΨ are convex functions of τ , giving free enthalpies Gˆ concave functions of pressures P. Specific free energy of the liquid phase Ψˆl = τlΨl involves a linear function of temperature −d τc (Tc − T ) which has a physical significance because the specific free energy of the vapor phase does not involve the same linear function, see Remark 13.12. In a cloud, relationships (13.43) and (13.44) give 1 1 − ) + d(T − Tc ) = 0, τv τl τv kT ln + d τc (T − Tc ) = 0. τl

− kT (

13.12 A Schematic Example in Case There Is No Air

257

Fig. 13.1 The schematic diagram. The isopressure P curve in red and the isospecific pressure p in black. They are the liquid Pl , pl pressure and specific pressure when specific volume and temperature are low and the vapor Pv , pv pressure and specific pressure when specific volume and temperature are large. At phase change temperature pressure and specific pressure are constant and specific volume is between τˆl (T ), the blue curve, and τˆv (T ) the green curve. These two schematic curves have a non smooth behaviour in the neighbourhood of critical temperature Tc. . They may be smoothened assuming more sophisticated free energies (see Remark 13.16 and Fig. 13.2)

Fig. 13.2 Functions τˆl (T ) and τˆv (T ) with smooth evolution in the neighbourhood of τc for liquid free energy (13.53)

258

13 Clouds: Mixture of Air, Vapor and Liquid Water

These equations give the two functions τˆl (T ) and τˆv (T ) satisfying

τˆl (T ) < τˆv (T ), lim

T →Tc ,T ≤Tc

τˆl (T ) =

lim

T →Tc ,T ≤Tc

τˆv (T ) = τc .

Functions τˆl (T ) and τˆv (T ) define the domain where there is a mixture of liquid and vapor in diagram of Fig. 12.3. We get

τˆl (T ) = Z=

τc expZ − 1 τc , τˆv (T ) = (exp Z − 1), Z exp Z Z dτc (Tc − T ), kT

see Fig. 12.3. In the neighbourhood of Tc , we have Z d τˆl (T ) τc (1 − ) = τc − τc (Tc − T ), 2 2kTc Z d τˆv (T ) τc (1 + ) = τc + τc (Tc − T ). 2 2kTc Remark 13.16. The two functions τˆl (T ) and τˆv (T ) are not tangent for T = Tc as in diagram of Fig. 12.3. To have this property free energy Ψl (τl , T ) has to be upgraded with a more sophisticated formula, for instance ⎧ (τc −τl )1/2 ln τ ⎪ (T − Tc ), i f τc − τl ≥ 0, T − Tc ≤ 0, ⎨ −kT τl l + d τl (13.53) Ψl (τl , T ) = I+ (τc − τl ), i f τc − τl < 0, ⎪ ⎩ −kT lnτlτl , i f τc − τl ≥ 0, T − Tc ≥ 0, valid for any T and τl . Indicator function I+ (τc − τl ) keeps τl lower than τc . Resulting functions τˆl (T ) and τˆv (T ) are shown in Fig. 13.2 with smooth evolution in the neighbourhood of τc . It is to be noted that function τl → Ψˆl (τl , T ) = τlΨl is convex and function T → Ψl (τl , T ) is concave. Experimental results may be found in [147], on the behaviour in the neighbourhood of the critical point. Pressures pw and P are: • In the liquid pl = −kT (ln τl − 1) + τcd(T − Tc ), Pl =

kT + d(T − Tc ). τl

(13.54)

13.12 A Schematic Example in Case There Is No Air

259

• In the vapor pv = −kT (ln τv − 1), Pv =

(13.55)

kT . τv

• They are equal in the mixture due to relationship (13.44) or (13.47), (Pa = 0 because there is no air). The isopressure and isospecific pressure curves T versus specific volume τ T =

pl + τc dTc τl (Pl + dTc ) ,T= , k + τl d k(1 − ln τl ) + τc d

T =

Pv pv τv , T = , k k(1 − ln τv )

have the behaviour shown in diagram of Fig. 12.3, see Fig. 13.1. In a liquid vapor mixture, the pressure and specific pressure depend only on temperature whereas in liquid phase and in vapor phase they depend on both temperature and specific volume, see Fig. 13.1. Remark 13.17. The schematic liquid and vapor pressures and specific pressures are equal at critical temperature Tc .

13.12.1 The Specific Entropies and the Specific Phase Change Latent Heat The specific entropies are

∂ τlΨl = k ln τl − d (τc − τl ) , ∂T ∂ τvΨv = k ln τv . sˆv (T, τv ) = − ∂T sˆl (T, τl ) = −

The specific latent heat liquid water vapor phase change latent heat at temperature T is   τv (13.56) T (sˆv (T, τv ) − sˆl (T, τl ) = T k ln + d (τc − τl ) ≥ 0. τl It is non negative as required, because τv > τc > τl .The specific latent heat, in agreement with experiments, is null for T = Tc because τˆv (Tc ) = τˆl (Tc ) = τc .

260

13 Clouds: Mixture of Air, Vapor and Liquid Water

Let us note that the predictive theory depends on four parameters, Tc , τc , k and d but they are sufficient for the theory to account for the basic phenomena.

13.13 Case Where Vapor, Liquid Water and Air Velocities Are Equal We assume Uv = Ul = Ua = U. We add the three linear momentum macroscopic equations and get (

βv βv 1 − βv dU = + + ) τa τv τl dt

−2 div(k grad βv ⊗ grad βv )   ∂ (τvΨv ) ∂ (τlΨl ) ∂ (τaΨa ) + grad βv (τv , T ) + (1 − βv) (τl , T ) + βv (τa , T ) ∂ τv ∂ τl ∂ τa +(1 − βv) {(λl + μl ) grad div U + μl Δ U} +βv {(λa + λv + μa + μv ) grad div U + (μa + μv ) Δ U} +faext + flext + fvext , d (sv + sl + sa ) + (sv + sl + sa ) div U − λ Δ ln T = D + Rext , dt       1 1 1 + Ψv (τv , T ) − pw − Ψl (τl , T ) − pw Ψa (τa , T ) − pa τa τv τl +Breac − 2kβ Δ βv = 0, Breac ∈ ∂ I(βv ), −pw = βv τv2

∂Ψv ∂Ψl (τv , T ) + (1 − βv)τl2 (τl , T ), ∂ τv ∂ τl

∂Ψv ∂Ψl (τv , T ) − (1 − βv)τl2 (τl , T ), ∂ τv ∂ τl   ∂Ψa 1 0 = βv (τv , T ) + pa 2 , ∂ τa τa

−pw (2βv − 1) = βv τv2

βv 1 − βv d β v 1 − βv ( + )+( + ) div U = 0, dt τv τl τv τl βv d βv ( ) + div U = 0. dt τa τa There are 10 equations and 10 unknowns U, βv , τv , τl , τa , pw , pa and T .

13.14 Case Where Air Velocity Is Null, Vapor and Liquid Velocities

261

13.14 Case Where Air Velocity Is Null, Vapor and Liquid Velocities Are Equal and Small We have Uv = Ul = U, Ua = 0. The equations of macroscopic motion of vapor and liquid of (13.30) give 1 dU = −2 div(k grad βv ⊗ grad βv ) − grad P˜ τ dt − (βv kva + βl kla ) U +βv {(λv + μv ) grad div U + μv Δ U} + (1 − βv) {(λl + μl ) grad div U + μl Δ U} +fvext + flext , with

1 βv 1 − β v = + , τ τv τl

and     1 1 ˜ P = −βv Ψv (τv , T ) − pw − (1 − βv) Ψl (τl , T ) − pw , τv τl

(13.57)

where P˜ is the vapor liquid partial pressure in the air vapor liquid mixture. It is the total pressure Pcloud in case there is no air. We assume the pressures and the friction forces between the air and the water are the leading terms of this equation of motion. The other forces, the acceleration forces, the exterior forces and the stresses which are not pressures, are negligible compared to them. By keeping only the leading terms in this equation of motion, we get ˜ (βv kva + βl kla ) U = − grad P. For the sake of simplicity, we assume also kva = kla = k p , and we have U=−

1 ˜ grad P. kp

(13.58)

This relationship means that vapor and liquid water flow toward low pressure. Note that this type of constitutive law is rather common (see for instance, relationship (4.29)). The mass balance (13.34) gives d dt

  1 1 + div U = 0. τ τ

(13.59)

262

13 Clouds: Mixture of Air, Vapor and Liquid Water

The air mass balance (13.4) becomes

∂ ∂t



βv τa

 = 0.

(13.60)

The other equations are 

 ∂Ψv 1 = βv 0 = βv + pw 2 ∂ τv τa   ∂Ψl 1 = (1 − βv) + pw 2 , ∂ τl τl       1 1 1 Ψa − pa + Ψv − pw − Ψl − pw τa τv τl

(13.61)

+Breac − 2kβ Δ βv = 0, Breac ∈ ∂ I(βv ),

(13.62)

dsv dsl ∂ sa + + + (sv + sl ) div U − λ Δ ln T = D + Rext . dt dt ∂t

(13.63)

∂Ψa 1 + pa 2 ∂ τa τa





˜ U, pa , pw , τv , τl , τa , βv and T . There are 9 equations, The 9 unknowns are P, (13.57)–(13.63) to find them.

13.14.1 The Example In the schematic example we have       1 1 1 + Ψv − pw − Ψl − pw δ = Ψa − pa τa τv τl   kT rT kT + + d(Tc − T ) . =− − τa τv τl Quantity δ governs the phase change. In an homogeneous mixture, i.e., Δ βv = 0, we have: • If δ > 0, βv = 0, there is only liquid water. Conditions of Sect. 13.11 apply and we have pl (τl , T ) > pˆw (T ) ⇔ τl < τˆl (T, τa = ∞) ⇔ Pcloud > Pˆl (T ), where specific pressure pˆw is the liquid–vapor mixture specific pressure pˆw (T ) = pˆw (T, τa = ∞), Pcloud = Pl and Pˆl (T ) is the liquid pressure in the vapor liquid water mixture at temperature T .

13.14 Case Where Air Velocity Is Null, Vapor and Liquid Velocities

263

• If δ < 0, βv = 1, there is only vapor and air. Conditions of Sect. 13.11 apply and we have pv (τv , T ) < pˆ w (T, τa ) ⇔ τv > τˆv (T, τa ) ⇔ Pv < Pˆv (T, τa ) ⇔ Pcloud < Pˆv (T, τa ) + Pa(T, τa ), where specific pressure pˆ w is the air liquid vapor mixture specific pressure pˆw (T, τa ) and Pˆv (T, τa ) is the vapor pressure in the air vapor liquid water mixture at temperature T with air specific volume τa . • If δ = 0, 0 ≤ βv ≤ 1, there is a mixture of air, liquid water and vapor with conditions given in Sect. 13.11

τl = τˆl (T, τa ), τv = τˆv (T, τa ), pv (τˆv , T ) = pl (τˆl , T ) = pˆw (T, τa ), Pcloud = Pˆv (T, τa ) + Pa (T, τa ) + Pˆl (T, τa ), Pˆv (T, τa ) + Pa (T, τa ) = Pˆl (T, τa ). Remark 13.18. The preceding results based on the potentials G may be obtained with the free energies. For instance, condition δ > 0, is     1 1 > Ψl (τl , T ) − pl (τl , T ) , inf Ψv (τv , T ) − pl (τl , T ) τv τv τl or

    1 1 Ψv (τ¯ (τl ), T ) − pl (τl , T ) > Ψl (τl , T ) − pl (τl , T ) , τ¯ τl

with pl (τl , T ) = −τl2

∂Ψl (τl , T ), ∂ τl

pl (τl , T ) = −τ¯v2

∂Ψv (τ¯v , T ). ∂ τv

and τ¯v (τl , T ) defined by

Function  τl → Ψv (τ¯v (τl , T ), T ) − pl (τl , T )

 1 τ¯v (τl , T )   1 = δ¯ (τl , T ), − Ψl (τl , T ) − pl (τl , T ) τl

is decreasing because (see Remark 13.11)

264

13 Clouds: Mixture of Air, Vapor and Liquid Water

1 ∂ δ¯ ∂ pl 1 =− ( − ) < 0. ∂ τl ∂ τl τ¯v (τl , T ) τl It is equal to 0 for τl = τˆl (T ) = τˆl (T, τa = ∞). Thus δ > 0 is equivalent to τl < τˆl (T ).

13.14.2 The Equations When the Temperature Is Known Assuming temperature is known, the equations are     1 1 ˜ P = −βv Ψv (τv , T ) − pw − (1 − βv) Ψl (τl , T ) − pw , τv τl

∂ βv ( ) = 0, ∂ t τa 1 d 1 ( )− Δ P˜ = 0, dt τ k pτ       1 1 1 Ψv (τv , T ) − pw + Ψa (τa , T ) − pa − Ψl (τl , T ) − pw τv τa τl + Breac − 2kβ Δ βv = 0, Breac ∈ ∂ I(βv ),       ∂Ψv 1 ∂Ψl 1 ∂Ψa 1 + pw 2 = (1 − βv) + p w 2 = βv + pa 2 , 0 = βv ∂ τv τv ∂ τl ∂ τa τa τl (13.64) 1 β v 1 − βv = + . τ τv τl ˜ pw , pa , τv , τl , τa , τ and βv . The two first relationships of The 8 unknowns are P, (13.64) are equivalent to − pw = βv τv2 2βv (1 − βv)τv2

∂Ψv ∂Ψl (τv , T ) + (1 − βv)τl2 (τl , T ), ∂ τv ∂ τl

∂Ψv ∂Ψl (τv , T ) = 2βv (1 − βv)τl2 (τl , T ). ∂ τv ∂ τl

13.14.2.1 An Other Formulation of the Equations For the sake of simplicity, let us assume there is no air. We may compute P˜ with relationships (13.64)

13.15 The Temperature Is Larger than Critical Temperature Tc

265

    1 1 ˜ P = −βv Ψv (τv , T ) − pw − (1 − βv) Ψl (τl , T ) − pw τv τl   ∂ (τvΨv ) ∂ (τlΨl ) (τv , T ) + βl (τl , T ) . = − βv ∂ τv ∂ τl We get the system of partial differential equations

∂Ψv ∂Ψl (τv , T ) = 2βv (1 − βv)τl2 (τl , T ), ∂ τv ∂ τl   1 d 1 ∂ (τvΨv ) ∂ (τlΨl ) (τv , T ) + βl (τl , T ) = 0, ( )+ Δ βv dt τ kpτ ∂ τv ∂ τl

2βv (1 − βv)τv2

(13.65)

1 βv 1 − βv = + , τ τv τl

∂ (τvΨv ) ∂ (τlΨl ) (τv , T ) − (τl , T ) + Breac − 2kβ Δ βv = 0, ∂ τv ∂ τl

(13.66)

Breac ∈ ∂ I(βv ),

(13.67)

for the 4 unknowns τv , τl , τ and βv .

13.15 The Temperature Is Larger than Critical Temperature Tc When temperature T tends towards Tc with T < Tc , the physical properties of the liquid phase and of the vapor phase become similar. When temperature is larger or equal to Tc they are the same. Let us note that the gas resulting from the liquid phase can mix with the air. Thus in case we assume a mixture of the three phases, we have to verify that there is no non miscible phase with the air phase, i.e., the liquid phase tends to have null volume fraction. In order to have the same mechanical properties when T ≥ Tc , we assume lim

T →Tc ,T Tc , constitutive laws (13.38) give ∂ ∂ (τv , T ) = τl2 (τl , T ). −pw = τv2 ∂ τv ∂ τl It results

τv = τl = τ .

266

13 Clouds: Mixture of Air, Vapor and Liquid Water

Thus the two water phases have same density and temperature. But they have different volume fractions βv and βl . Assuming air is an ideal gas, we have       1 1 1 Ψa (τa , T ) − pa + Ψv (τv , T ) − pw − Ψl (τl , T ) − pw τa τv τl =−

rT < 0. τa

(13.68)

Equation (13.27) which governs the evolution of the mixture −

rT + Breac − 2kβ Δ βv = 0, τa

tends to impose βv = 1 because −rT /τa is negative. It appears the presence of air favors the vapor phase by having rT /τa = Pa large. In case there is no air, the volume fractions do not intervene on the evolution. The results are the same whatever their values.

13.15.1 An Homogeneous Cloud In an homogeneous domain, relationships (13.52) and (13.68) give βv = 1. There is only vapor and air with equations of motion and entropy balance 1 dv Uv = −kva (Uv − Ua ) + fvext , τ dt 1 da Ua = kva (Uv − Ua ) + faext , τa dt pw = −τ 2

∂ ∂Ψ (τ , T ), pa = −τa2 (τv , T ), ∂τ ∂ τa

dv sv da sa + = D + Rext . dt dt

(13.69)

(13.70) (13.71)

Note that the dissipative terms kva (Uv − Ua ) in (13.69) and (13.70) tend to equalize the air and gas velocity in the cloud.

13.15.2 The Example: The Different Temperature-Specific Volume Domains The free energies defined for any temperature and specific volume are

Ψl (τl , T ) = −kT

ln τl τc − τl −d (pn(T − Tc )) , τl τl

13.15 The Temperature Is Larger than Critical Temperature Tc

Ψv (τv , T ) = −kT

267

ln τv , τv

where function pn {x} = sup(0, −x) is the negative part function. It results (τ , T ) = −kT

ln τ . τ

(13.72)

13.15.2.1 The Schematic Liquid–Vapor Domains Depending on the temperature and on vapor specific volume, we have four schematic domains: • When temperature is lower than critical temperature Tc , vapor and liquid may coexist. The two schematic free energies are

Ψl (τl , T ) = −kT

τc − τl ln τl +d (T − Tc ), τl τl

Ψv (τv , T ) = −kT

ln τv . τv

(13.73)

This domain, T ≤ Tc , is divided into three domains depending on the specific volume: 1. If the specific volume is low, τ < τˆl (T ), there is only liquid. 2. If the specific volume is in between τˆl (T ) and τˆv (T ) there is a mixture of liquid and vapor. 3. If the specific domain is large, τ > τˆv (T ), there is only vapor. • When temperature is larger than critical temperature Tc , there is only vapor if air is present. It is no longer possible to discriminate vapor and liquid when there is no air. They have the same free energy F (τ , T ) = −kT

ln τ , τ

which is the free energy of an ideal gas. The only difference between vapor and liquid is that air is not miscible with liquid. Remark 13.19. Functions T → Ψ (τ , T ) are concave functions for T = Tc because T → − (τc − τl ) pn(T − Tc ) is concave because τc − τl > 0. Remark 13.20. Mechanics requires that function T → Ψ (τ , T ) is concave but does not require function 1/τ → Ψ (τ , T ) is convex. This property is satisfied by the free energies we have chosen. It results the analysis of the evolution is rather easy and simple. In case function 1/τ → Ψ (τ , T ) is not convex, definition of potential G is questionable and the analysis may be difficult.

268

13 Clouds: Mixture of Air, Vapor and Liquid Water

Remark 13.21. A part of physics is not described: the case ice is present in the cloud. Ice appears when temperature is low and specific volume is small. In a predictive theory introducing the possibility of ice, a solid volume fraction intervene, βs , with βs + βl + βv = 1, assuming there are no voids and no interpenetration of the phases. The same rules provide the equations of the predictive theory. There is the possibility to have the three phases present at the triple point (see diagram of Fig. 12.1).

13.16 Solid–Solid and Solid–Gas Phase Changes A gas, for instance hydrogen, circulating within a porous metal may be adsorbed by the metal acting as a gasholder. The gas-metal mixture is a mixture of two solid solutions, [154]. The solid solutions mixture composition varies involving solid– solid and solid–gas phase changes. Experiments show that the temperature and the gas pressure are the important quantities to describe this solid–solid phase change which is important when focusing on the amount of gas which is adsorbed, [154]. Assuming the macroscopic approximation that the two phases can coexist at each point together with gas and metal, a predictive theory together with analytical results are described in [45]. The state quantities are E = (T, τ , β , grad β ) for the metal and the solid solutions and, Eg = (T, τg ) for the gas, T is the temperature, τ the solid solution specific volume, β the volume fraction of a solid solution, 1 − β being the volume fraction of the other solid solution and τg is the specific gas volume.

Chapter 14

Conclusion

Continuum and discontinuum mechanics give predictive theories to investigate at the engineering level, i.e., the macroscopic level, numerous phase change phenomena occurring during smooth or violent evolutions. The important elements of the theories are the choice of: 1. The work of the interior forces 2. The free energy 3. The pseudopotential of dissipation The choice of the work of the internal forces gives the sophistication and the scope of the theories. The choices of the free energy and of the pseudopotential of dissipation which gather all the physical properties, is guided both by theory and experiment. Because experiments are difficult to perform, expensive end time consuming, the imagination and skill of the engineer and scientist is important at this step of the construction of the predictive model. The theory has to have few parameters to be useful. Note that this is the case for the cloud theory, Chap. 13, involving only four parameters. Other elements of the modelling work have not been mentioned: 1. The mathematical analysis: it is important to know that the equations are coherent in terms of mathematics. A computer program gives always numerical answers even to problems which have no solution. It is difficult to trust them. 2. The numerical approximations. 3. The measurements of the physical parameters. The theories may be upgraded or downgraded to apply to macroscopic engineering problems coupling mechanical and thermal actions.

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8 14, © Springer-Verlag Berlin Heidelberg 2012

269

•

Appendix A

Some Elements of Convex Analysis

A.1 Convex Sets Let C be a set of linear space V . This set is convex if Definition A.1. ∀x ∈ C, ∀y ∈ C, ∀θ ∈ ]0, 1[, point θ x + (1 − θ )y ∈ C. Segment [0, 1] of space V = R is convex. The interior of a circle is a convex set of V = R2 . On the contrary the exterior of a circle is not convex. Convex sets are useful in mechanics to describe numerous and various properties. For instance, the possible positions of a soccer ball above the football field is a convex set: it is C = {x = (xi ), i = 1, 3 |xi ∈ R, x3 ≥ 0} .

A.2 Convex Functions Let R = R ∪ {+∞} where the regular addition is completed by the rules ∀a ∈ R, a + (+∞) = +∞, +∞ + (+∞) = +∞. Multiplication by positive numbers is completed by ∀a ∈ R, a > 0, a × (+∞) = +∞. In this context, it is forbidden to multiply either by 0 or by negative numbers. Let f an application from linear space V into R. This function is convex if

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8, © Springer-Verlag Berlin Heidelberg 2012

271

272

A Some Elements of Convex Analysis

Definition A.2. ∀x ∈ V, ∀y ∈ V, ∀θ ∈ ]0, 1[, f (θ x + (1 − θ )y) ≤ θ f (x) + (1 − θ ) f (y). Remark A.1. Function f is actually multiplied by positive numbers because θ ∈ ]0, 1[.

A.2.1 Examples of Convex Functions It is easy to prove that functions f p where V = R x ∈ R → f p (x) =

1 p |x| , with p ∈ R, p ≥ 1, p

are convex. Let function I from V = R into R be defined by I(x) = 0, i f x ∈ [0, 1] , I(x) = +∞, i f x ∈ / [0, 1]. This function is convex. It is called the indicator function of segment [0, 1] (Fig. A.2). More generally, we denote indicator function of set C ⊂ V , function IC defined by IC (x) = 0, i f x ∈ C, / C. IC (x) = +∞, i f x ∈ It is easy to connect the convex function and convex set notions. It is shown that Theorem A.1. A convex C ⊂ V is convex if and only if its indicator function IC is convex. Indicator functions may seem a little bit strange. In the sequel it is to be seen in the examples that they are productive tools in mechanics for dealing with internal constraints relating mechanical quantities. Three other indicator functions are useful for V = R. They are the indicator functions I+ , I− and I0 of the sets of the non negative and non positive numbers and of the origin: I+ (x) = 0, i f x ≥ 0, I+ (x) = +∞, i f x < 0, I− (x) = 0, i f x ≤ 0, I− (x) = +∞, i f x > 0,

A.3 Linear Spaces in Duality

273

and I0 (0) = 0, I0 (x) = +∞, i f x = 0.

A.3 Linear Spaces in Duality Definition A.3. Two linear spaces V and V ∗ are in duality if there exist a bilinear form ·, · defined on V × V ∗ such that f or any x ∈ V, x = 0, there exists y∗ ∈ V ∗ , such that x, y∗ = 0; f or any y∗ ∈ V ∗ , y∗ = 0, there exists x ∈ V, such that x, y∗ = 0.

A.3.1 Examples of Linear Spaces in Duality Spaces V = R, V ∗ = R are in duality with the bilinear form which is the usual product x, y = x · y ;

(A.1)

V = Rn , V ∗ = Rn are in duality with the bilinear form which is the usual scalar product i=n

x, y = x · y = ∑ xi yi ,

(A.2)

i=1

where x = (xi ) and y = (yi ) are vectors of Rn , the coordinates of which are xi and yi ; V = S, V ∗ = S where S is the linear space of symmetric matrices 3 × 3, are in duality with the bilinear form e ∈V, s ∈ V ∗ , e, s = e : s =

i, j=3

∑

ei, j si, j = ei, j si, j

i, j=1

= e11 s11 + e22s22 + e33s33 + 2e12s12 + 2e13s13 + 2e23s23 , (A.3) where we use the Einstein summation rule. Be careful, linear spaces V = S and V ∗ = S are also in duality with the bilinear form

274

A Some Elements of Convex Analysis

e, s = e11 s11 + e22s22 + e33s33 + e12s12 + e13 s13 + e23 s23 , which is different from the preceding one. Let us give two more examples. Linear space of the velocities U in a domain Ω of R3    V = U(x) U ∈ L2 (Ω ) , is in duality with the linear space of the forces f applied to the points of the domain    V ∗ = f(x) f ∈ L2 (Ω ) , with bilinear form U, f =

 Ω

U(x) · f(x)d Ω ,

which is the power of the force applied to the domain. The linear space of the strain rates    V = D = (Di j (x)) Di j = D ji , Di j ∈ L2 (Ω ) , is in duality with the stresses linear space    V ∗ = σ = (σi j (x)) σi j = σ ji , σi j ∈ L2 (Ω ) , with bilinear form D, σ =

 Ω

σ (x) : D(x)d Ω =

 Ω

σi j (x)Di j (x)d Ω ,

which is the power of the stresses. This bilinear form is used in the definition of the power of the interior forces, see Sect. 3.1.

A.4 Subgradients and Subdifferential Set of Convex Functions Convex function f p given above, is differentiable for p > 1 but it is not for p = 1 where it is equal to the absolute value function x → |x|, which has no derivative at the origin. In the same way, indicator function I is not differentiable because its value is +∞ at some points and it has no derivative at points x = 0 and x = 1. Let us recall that for smooth convex functions of one variable, the derivative is an increasing function and this property yields 

 df df (y) − (z) (y − z) ≥ 0. dx dx

(A.4)

A.4 Subgradients and Subdifferential Set of Convex Functions

275

Fig. A.1 Convex function f has not a derivative at point A. It has generalized derivatives: the slopes of the lines which pass at point A and are under the curve representing function f . These slopes are the sub-gradients which constitute the subdifferential set

Thus it seems that we have to loose all the calculus properties related to derivatives. Fortunately, this is not the case. Indeed, it is possible to define generalized derivatives and keep a large amount of properties related to derivatives. Consider function f shown in Fig. A.1. It is convex but it is not differentiable because it has no derivative at point A. At a point where the function has a derivative, the curve is everywhere above the tangent. At point A, there exist several lines which have this properly. The slopes of these lines are the subgradients which generalize the derivative: Definition A.4. Let convex function f defined on V in duality with V ∗ . A subgradient of f at point x ∈ V is an element x∗ ∈ V ∗ which satisfies ∀z ∈ V, z − x, x∗ + f (x) ≤ f (z).

(A.5)

The set of the x∗ which satisfy (A.5) is the subdifferential set f at point x, denoted ∂ f (x). Remark A.2. The subgradient depends on f but depends also on the bilinear form ·, · .

A.4.1 Two Properties of the Subdifferential Set The subdifferential set keeps usual properties of the derivative: a function is not differentiable where its value is +∞ and a differentiable convex function satisfies relationship (A.4). For a convex function first property becomes: Theorem A.2. Let convex function f = +∞. If this function is subdifferentiable at / then it is finite at that point: f (x) < +∞. point x, i.e., if ∂ f (x) = 0,

276

A Some Elements of Convex Analysis

Fig. A.2 Indicator function of segment [0, 1]. Its value I(x) is 0, if 0 ≤ x ≤ 1, and +∞, if either x < 0 or x > 1. The subgradients at points 0 and 1 are the slopes of the lines which are under the curve representing the function and in contact with the curve at points 0 or 1

Proof. Let us assume f is subdifferentiable at point x: let y∗ ∈ ∂ f (x). Let us reason ab absurdo and assume that f is not finite at that point: f (x) = +∞. Let us write relationship (A.5) at a point z where f (z) < +∞. Such a point exists because f = +∞. Then we have +∞ = z − x, y∗ + f (x) ≤ f (z). We deduce that +∞ = f (z). Which is contradictory with the assumption. Let us note that if the only point z where f (z) < +∞ is x itself, then f (x) < +∞ for f not to be identical to +∞. 

This theorem applies in numerous constitutive laws where we have the relationship B ∈ ∂ I(β ): because ∂ I(β ) is not empty, we have I(β ) < +∞ which implies that I(β ) = 0 and 0 ≤ β ≤ 1, see Fig. A.2. Thus relationship B ∈ ∂ I(β ) implies that quantity β which may be a phase volume fraction is actually in between 0 and 1. Now let us prove that relationship (A.4) is satisfied in some sense by the subdifferential set : Theorem A.3. Let f a function convex. We have ∀y∗ ∈ ∂ f (y), ∀z∗ ∈ ∂ f (z), y − z, y∗ − z∗ ≥ 0. Proof. It is sufficient to write relationship (A.5) at points x and y.



Remark A.3. It is said that the subdifferentiation operator is a monotone operator.

A.4.2 Examples of Subdifferential Sets Let us begin by the subdifferential set of indicator function I of interval [0, 1]. It is easy to see that the subdifferential set is (Figs. A.2 and A.3)

∂ I(0) = R− ; ∂ I(1) = R+ ; i f x ∈ ]0, 1[ , ∂ I(x) = {0} ; / if x ∈ / [0, 1], ∂ I(x) = 0.

A.4 Subgradients and Subdifferential Set of Convex Functions

277

Fig. A.3 On the left, subdifferential set ∂ I+ of indicator function of the set of the positive numbers R+ : ∂ I+ (0) = R− , ∂ I+ (x) = {0} for x > 0 and ∂ I+ (x) = 0/ for x < 0. On the right, subdifferential set ∂ I of indicator function I of segment [0, 1]

Fig. A.4 Vector B ∈ ∂ IC (β ) is normal to triangle K. Vectors B at vertices A and B.and on side AB of triangle K

The subdifferential of indicator function I+ is also easily computed. It is shown in Fig. A.3. Let us consider triangle K with vertices O, A and B. It is a convex set of R2 (Fig. A.4). Let IK be its indicator function. Let us compute its subdifferential set when the bilinear form is the usual scalar product (A.2) of R2 . Then a subgradient of indicator function IK at point x is a vector B which satisfies (A.5) ∀y ∈ R2 , IK (y) ≥ IK (x) + (y − x) · B. We deduce that: If x ∈ / K, there exists no vector B which satisfies the previous relationship (apply Theorem A.2). If x ∈ K, the previous relationship gives, ∀y ∈ K, 0 ≥ (y − x) · B. This relationship proves that vector B is normal to convex set K (Fig. A.4).

278

A Some Elements of Convex Analysis

A.5 Dual Functions Let spaces V and V ∗ in duality with bilinear form ., . . Let f a convex function of V into R = R ∪ {+∞}, the dual function f ∗ of f is a function of V ∗ into R defined by f ∗ (y∗ ) = sup{x, y∗ − f (x) |x ∈ V }. It is possible to prove Theorem A.4. Dual function f ∗ is convex. If function f is subdifferentiable at point x, the properties y∗ ∈ ∂ f (x), x ∈ ∂ f ∗ (y∗ ), and f (x) + f ∗ (y∗ ) =< x, y∗ >,

(A.6)

are equivalent. As for an example, let us compute dual function of the indicator function I of segment [0, 1] with V ∗ = R and the bilinear form being the usual multiplication x, y = xy. Dual function is defined by I ∗ (y) = sup{xy − I(x) |x ∈ V = R } = sup{xy |x ∈ [0, 1] }. It is called the support function of segment [0, 1]. It easy to get I ∗ (y) = y, i f y ≥ 0, and I ∗ (y) = 0, i f y ≤ 0. Function I ∗ is the positive part function defined by pp(y) = sup {y, 0} .

A.5.1 The Internal Energy, Potential G and Free Enthalpy Free energy depends on temperature T and on other state quantities η , Ψ (T, η ). It is a concave function of temperature. Dual function of its opposite function, which is convex (−Ψ )∗ (y, η ) = sup{xy − (−Ψ )(x, η ) |x ∈ V = R }, defines the internal energy depending on entropy s and on η by relationship e(s, η ) = (−Ψ )∗ (s, η ). It results also the equivalent relationships

A.6 Concave Functions

279

T ∈ ∂ e(s, η ), s ∈ −∂Ψ (T, η ), and e(s, η ) = sT + Ψ (T, η ), where the subdifferential sets are with respect to s and T , η being a parameter, see also Sect. 3.2.3. Note that ∀x, ∀y, e(y, η ) − Ψ (x, η ) ≥ xy. An other example of dual function is given in Sect. 13.10.1, where a new potential G is defined with dual function of volumic free energy Ψ with respect to density 1/τ . We define function Ψ ∗ (p, T ) by  1 Ψ ∗ (p, T ) = sup p − Ψ (x, T ) |x ∈ V = R , x x where function 1/τ → Ψ (τ , T ) is convex and potential G(p, T ) by G(p, T ) = −Ψ ∗ (p, T ), with 1 ∂Ψ ∗ ∂G ∂Ψ ∂Ψ 1 . G(p, T ) = −p + Ψ (τ , T ) ⇔ = =− ⇔ p= = −τ 2 τ τ ∂p ∂p ∂ (1/τ ) ∂τ Remark A.4. Note that function p → Ψ ∗ (p, T ) is the dual function of function y → Ψ (1/y, T ). To avoid too many notations, we do not denote this function. Free enthalpy is defined with specific free energy Ψˆ = τΨ   ˆ T ) = Ψˆ ∗ (−P, T ) = sup −Px − Ψˆ (x, T ) , −G(P, x

with

ˆ ˆ ˆ T ) = Pτ + Ψˆ (τ , T ) ⇔ τ = ∂ G ⇔ P = − ∂ Ψ . G(P, ∂P ∂τ Precise presentation of convex analysis with applications is given either in books by Jean Jacques Moreau, [172] and by Ivar Ekeland and Roger Temam, [90] or by Bernard Nayroles, [174].

A.6 Concave Functions A function f is concave if its opposite − f is convex. For concave function, uppergradient, instead of subgradient, are defined. They satisfy ∀z ∈ V, z − x, x∗ + f (x) ≥ f (z).

280

A Some Elements of Convex Analysis

We denote ∂ˆ f (x), upperdifferential set, the set of the uppergradient at point x. It is easy to prove that ∂ˆ f (x) = −∂ (− f )(x).

A.6.1 Example of Concave Functions The free energy Ψ (T, ε , β , grad β ) is a concave function of temperature T , because it is the dual function of the opposite of internal energy e(s, ε , β , grad β ) which is a convex function of entropy s, [132]. It results the heat capacity −

∂ 2Ψ , ∂T2

is non negative. Remark A.5. If the internal constraint (3.21) is taken into account by

Ψ (T, η ) = Ψ (T, η ) − I+(T ), the free energy Ψ (T, η ) is still a concave function of T , [112]. ˆ T ) depending on pressure P and temperature T , is concave with Potential G(P, respect to T and concave with respect to P.

Appendix B

The Small Perturbation Assumption

Let us consider the position x at time t of a material point whose position at time 0 is a. We let x(a,t) = a + u(a,t),

(B.1)

where vector u(a,t) is the displacement vector. The set occupied by the material points at time 0 is the reference configuration, Ωa . The basic idea of the small perturbation assumption is to assume that the displacement u(a,t) = x(a,t) − a, is small compared to some reference displacement.

B.1 Assumptions on the Displacement and on Mechanical Quantities We assume ASSUMPTION 1. The function u(a,t) and its derivatives are of order η

i+ j

∂ u(a,t)



∂ t i ∂ a j ≤ cη , for any i,j. The parameter η is small with respect to 1 and c is a constant independent of i,j. Let us recall that function O(a,t, z) is such that O(a,t, z) = A, z→0 z lim

where A ∈ R. M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8, © Springer-Verlag Berlin Heidelberg 2012

281

282

B The Small Perturbation Assumption

Assumption 1 allows to compare derivatives of the displacement either in the reference configuration, Ωa or in the actual configuration, Ω (t) Proposition B.1. Let ua (a,t) = u(a + u(a,t),t) = u(x,t), then if assumption 1 is satisfied u(x,t) = ua (a,t) = u(a,t) + O(a,t, η 2 ,

∂u (x,t) = ∂x du (x,t) = dt

∂u ∂u ∂u (a,t) + O(a,t, η 2 )= (x,t) + O(a,t, η 2 ) = (a,t) + O(a,t, η 2 ), ∂x ∂a ∂a ∂ ua ∂u ∂u (a,t) = (x,t) + O(x,t, η 2 ) = (a,t) + O(a,t, η 2 ). ∂t ∂t ∂t

Proof. Let us prove for instance, the last relationships related to the velocity. We have du ∂ ua ∂u ∂u ∂u (x,t) = (a,t) = (x,t) + (x,t) (a,t) dt ∂t ∂t ∂x ∂t ∂u ∂u = (x,t) + O(x,t, η 2 ) = (a,t) + O(a,t, η 2 ), ∂t ∂t because

∂u (x,t) = ∂x ∂u (x,t) = ∂t

∂u (a,t) + O(a,t, η 2 ), ∂x ∂u (a,t) + O(a,t, η 2 ), ∂t 

due to assumption 1. Remark B.1. We may prove that

∂ i+ j u(a,t) ∂ i+ j ua (a,t) = + O(a,t, η 2 ). ∂ t i∂ a j ∂ ti∂ a j It results from Proposition B.1 that the displacement and its derivatives are given, up to an O(η 2 ) quantity, by the same function either in the reference configuration with position a or in the actual configuration with position x (the positions being related by relationship (B.1)). This property is important for the velocity U(x,t) = U(a,t) + O(a,t, η 2 ). It is also important for mechanical quantities related to displacement, for instance the deformation, computed with the displacement u(a,t) may be estimated with assumption 1. Consider gradient matrix F

B.1 Assumptions on the Displacement and on Mechanical Quantities

F(a,t)=

283

∂x ∂u (a,t) = (a,t) + 1 = H + 1, ∂a ∂a

where H is of order 1, H ≤ cη . We have G = F−1 = −H + 1 + O(η 2), C = FT F = 1 + H + HT + O(η 2 ) = 1 + 2ε + O(η 2 ),

∂a (x,t) = F−1 (x−1 (x,t)) = −H + 1 + O(η 2), ∂x where matrix

H + HT , 2 is the small deformation matrix. Proposition B.1 shows that F, G, H, ε have the same value up to an O(η 2 ) quantity, either in the reference configuration with position a or in the actual configuration with position x (the positions being related by relationship (B.1)). Assumption 1 and Proposition B.1 concerning the displacements are extended to other mechanical quantities with two slightly different assumptions, either

ε=

ASSUMPTION 2.1 Let a mechanical quantity be g(a,t) in the reference framework, Ωa , and f (x,t) in the actual framework, Ω (t). They are related by g(a,t) = f (a + u(a,t),t). It is assumed that functions g, f and their first and second derivatives are of order 1 with respect to η , or ASSUMPTION 2.2 Let a mechanical quantity be g(a,t) in the reference framework, Ωa , and f (x,t) in the actual framework, Ω (t). They are related by g(a,t) = f (a + u(a,t),t). It is assumed that functions g, f are of order 0 and their first and second derivatives are of order 1 with respect to η . The values of the mechanical quantities at points a and x related by (B.1) may be compared Proposition B.2. If assumptions 1 and either 2.1 or 2.2 are satisfied, functions g and f satisfy f (x,t) = g(x,t) + O(x,t, η 2 ) = g(a,t), g(a,t) = f (a,t) + O(a,t, η 2 ),

∂f (x,t) = ∂x df (x,t) = dt

∂g ∂g ∂f (a,t) + O(a,t, η 2 ) = (x,t) + O(x,t, η 2 ) = (a,t) + O(a,t, η 2 ), ∂a ∂a ∂x ∂g ∂f (a,t) = (x,t) + O(x,t, η 2 ) ∂t ∂t ∂f ∂g = (a,t) + O(a,t, η 2 ) = (a,t) + O(a,t, η 2 ). ∂t ∂t

284

B The Small Perturbation Assumption

Proof. We have   ∂f ∂x ∂f (x,t) (a,t) = (x,t) + O(η 2 ), ∂x ∂a ∂x  2  ∂g ∂g ∂ g ∂g (a,t) = (x,t) + (x,t) + O(η 2 ), (x,t) (−u(a,t)) + O(η 2 ) = 2 ∂a ∂a ∂a ∂a

∂g (a,t) = ∂a



because H ≤ cη due to assumption 1. We have also

∂g ∂f ∂u ∂ f df (x,t) = (a,t) = (x,t) + · , dt ∂t ∂t ∂t ∂ x which gives

df ∂g ∂f (x,t) = (a,t) = (x,t) + O(η 2 ), dt ∂t ∂t due to assumption 1. We have also  2  ∂f ∂f ∂ f (x,t) = (a,t) + (a,t) · u(a,t) + O(η 2 ), ∂t ∂t ∂ t∂ a  2  ∂g ∂g ∂ g (a,t) = (x,t) + (x,t) · (−u(a,t)) + O(η 2 ), ∂t ∂t ∂t∂ x

which gives

∂f ∂f ∂g ∂g (x,t) + O(η 2 ) = (a,t) + O(η 2) = (a,t) + O(η 2 ) = (x,t) + O(η 2 ). ∂t ∂t ∂t ∂t The consequence of the propositions is that any property, up to an O(η 2 ) quantity, is described by the same function either in the reference configuration with position a or in the actual configuration with position x, the positions being related by relationship (B.1).

B.1.1 The Surface Normal Vector Consider the surface vector 1 Na d Γa = FT Nx d Γx , J where J = det F = 1 +trε + O(η 2 )=1 + O(η ) and a = x−1 (x,t) are related by (B.1). It results Na d Γa = Nx d Γx + O(η ).

B.3 The Mass Balance

285

Note that the surface vector when multiplied by a mechanical quantity of order η is given by the same formula either in the reference configuration or in the actual configuration up to an O(η 2 ) quantity. Thus the boundary relationships have the same properties than the volume relationships.

B.2 Assumptions on the Exterior Forces: The Choice of the Small Perturbation Equations The small perturbation assumption involves two more assumptions on the exterior forces and on the choice of the equations to solve ASSUMPTION 3 The exterior forces and their first derivatives with respect to x and a are of order η . It results that the exterior forces are given by the same functions either of a,t or of x,t, up to an O(η 2 ) quantity. ˆ βˆ , grad βˆ and θˆ the solutions in reference configuration Ω a of Let us define u, the equations resulting from the balance laws and constitutive laws resulting from assumptions 1, 2 and 3 where the O(η 2 ) quantities are neglected. The following assumption compares the solutions of the different equations. ˆ βˆ , grad βˆ and θˆ are approximations of u, β , grad β ASSUMPTION 4 Functions u, and θ the solutions of the equations in Ω (t) where the O(η 2 ) quantities are not neglected, i.e. the regular equations. Remark B.2. Let us stress the subjective status of the small perturbation assumption. It is, like the choice of the state quantities, a choice of the engineer. ˆ βˆ , Remark B.3. It is to have a mathematical comparison of the approximations u, ˆ ˆ grad β and θ and the solutions u, β , grad β and θ . In most cases this comparison is completely out of reach due to mathematical difficulties and mostly to the fact that the constitutive laws are not clearly known outside the small perturbation assumption. Thus the difficulty is both mechanical and mathematical. A few results are available for rods, [8, 66, 192, 193].

B.3 The Mass Balance It is in the reference configuration 1 ρ (x(a,t),t) = ρa (a) = ρa (a)(1 − trε (a,t))+O(η 2 ), J where we assume functions u(x,t) satisfies Assumption 1 and ρ (x,t), ρa (a) satisfy Assumption 2.2. The mass balance is in the actual configuration

286

B The Small Perturbation Assumption

dρ (x,t) + ρ (x,t) div U(x,t) = 0, in Ω (t), dt or with Assumption 1 and Proposition B.2,

∂ρ (x,t) + ρa (x−1 (x,t)) div U(x,t) = 0 + O(η 2 ). ∂t This relationship is satisfied either in Ω (t) or in Ωa . The boundary condition −ρ (x,t)U(x,t) · Nx (x,t)d Γx (x,t) = m(x,t)d Γx (x,t), on ∂ Ω (t), where m(x,t) is the mass intake which is assumed to satisfy Assumption 3. We have −ρ (x(a,t),t)U(x(a,t),t) · Nx (x(a,t),t)dΓx (x(a,t),t) = m(x(a,t),t)dΓx (x(a,t),t) = −ρ (x(a,t),t)U(x(a,t),t) · (Na (a)dΓa (a) + O(η )), on ∂ Ωa , or with Propositions B.1 and B.2 −ρa U · N = m + O(η 2 ), either in ∂ Ω (t) or in ∂ Ω a . From Assumption 4, the small perturbation mass balance is

∂ ρˆ ˆ = 0, in Ωa , + ρa div U ∂t ˆ · N = m, on ∂ Ω a , − ρa U

(B.2) (B.3)

or ˆ = ρˆ , in Ωa , ρa (1 − trε (u)) −ρauˆ · N =

 t 0

md τ , on ∂ Ω a .

(B.4) (B.5)

In case the small displacement uˆ is known, (B.2) or (B.4) give density ρˆ , and (B.3) and (B.5) give the mass flux. In case the density depends on other quantities, relationships (B.2) or (B.4) are internal constraints between the state quantities or their velocities.

B.4 The Equations of Motion B.4.1 The Equation for the Macroscopic Motion The volume and surface forces, f and g, satisfy Assumption 3. Density ρ satisfies Assumption 2.2. We assume also the Cauchy stress Σ, the Piola-Kirchoff stress S

B.4 The Equations of Motion

287

and Boussinesq or Piola-Lagrange stress Π related by JGΣGT = S = GΠ, satisfy Assumption 2.1. This assumption is to result from the constitutive laws. The equation of motion in the actual configuration are dU = div Σ + f, in Ω (t), dt ΣNx = g, on ∂ Ω (t).

ρ

They are in the reference configuration

∂U = div Π + Jf, in Ωa , ∂t J ΠNa = g, on ∂ Ωa . F

ρa

It results the equations of motion

ρa

∂U = div Σ + f + O(η 2 ), either in Ωa or in Ω (t), ∂t ΣN = g + O(η 2 ), either on ∂ Ωa or on ∂ Ω (t),

because each quantity of the equations is given by the same function either in the actual or in the reference configuration up to an O(η 2 ) function. With Assumption 4, the small perturbation equation for the macroscopic motion is

ρa

ˆ ∂U ˆ + f, in Ωa , = div Σ ∂t ˆ = g, on ∂ Ω a . ΣN

B.4.2 The Equation for the Microscopic Motion: The Classical Case The equations for the microscopic motion in the actual configuration are − B + divH = A, in Ω (t), H · Nx = a, on ∂ Ω (t). They are in the reference configuration

(B.6)

288

B The Small Perturbation Assumption

− JB + divJHGT = JA, in Ωa , JHGT · Na =

J a, on ∂ Ωa . F

We assume that exterior works A and a satisfy Assumption 3 and that internal works B and H satisfy Assumption 2.1. It results from Proposition B.2 that the small perturbation equation for the microscopic motion is ˆ = A, in Ω a , − Bˆ + div H ˆ · N = a, on ∂ Ω a . H

(B.7)

The quantities which appear in this relationship are of order η .

B.4.3 The Equation for the Microscopic Motion: The Non Classical Case In this particular case, the assumptions are modified: the power of the order with respect to η is increased by 1 for the equation for the microscopic motion (thus it is η 2 ). For the other equations the order remains equal to η . The new assumptions are ASSUMPTION 2.1bis. Let a mechanical quantity which intervenes in the equation for the microscopic motion, to be g(a,t) in the reference framework and f (x,t) in the actual framework. They are related by g(a,t) = f (a + u(a,t),t). It is assumed that functions g, f and their first and second derivatives are of order η 2 . ASSUMPTION 3bis The external actions and their first derivatives with respect to x and a are of order η 2 . We assume that exterior works A and a satisfy Assumption 3bis and that internal works B and H satisfy Assumption 2.1bis. ˆ βˆ , grad βˆ and θˆ the solutions in reference configuration Ω a Let us define u, of the equations resulting from the balance laws and constitutive laws resulting from Assumptions 1, 2bis and 3bis where the O(η 3 ) quantities are neglected in the equation for the microscopic motion and the O(η 2 ) quantities are neglected in the other equations. The following assumption compares the solutions of the different equations. ˆ βˆ , grad βˆ and θˆ are approximations of u, β , ASSUMPTION 4bis Functions u, grad β and θ the solutions of the equations in Ω (t) where the O(η 3 ) quantities in the equation for the microscopic motion equation and the O(η 2 ) quantities are not neglected, i.e. the regular equations. The small perturbation equation for the microscopic motion are the same but the quantities are of order η 2 . This non classical situation is used in the damage theory where the elastic work

B.6 The Constitutive Laws

289

1 ˆ 2 + 2μe ε (u) ˆ : ε (u)}, ˆ {λe (trε (u)) 2 ˆ is is a damage source. This quantity is of order 2 when the small deformations ε (u) of order η .

B.5 The Entropy Balance It is in the actual configuration

ρ

ds 1 dβ d dβ Q · gradT + div Q = R + (Σd : D(U) + B + Hd · grad − ), in Ω (t), dt T dt dt T −Q · Nx = π , on ∂ Ω (t).

We assume T = T0 + θ , where θ satisfies Assumption 2.1 and T0 is of order 0, η 0 = 1, with respect to η , (T satisfies Assumption 2.2). We assume the external thermal actions R and π satisfy Assumption 3. We assume the interior forces satisfy Assumption 2.1 or 2.1bis, that the entropy flux vector satisfies Assumption 2.1 and that density ρ satisfies Assumption 2.2. It results dβ d dβ Q · gradT 1 d (Σ : D(U) + B + Hd · grad − ) = O(η 2 ). T dt dt T The small perturbation entropy balance is

ρ

∂ sˆ ˆ = R, in Ωa , + div Q ∂t ˆ · N = π , on ∂ Ω a . −Q

B.6 The Constitutive Laws They are, for instance, defined by the free energy Ψ (L, β , grada β ) and pseudopotential Φ (D(U), d β /dt, gradx d β /dt, gradx T ) where L = (FT F − I)/2 is the Green-Lagrange strain s=−

Σd = −Q =

∂Ψ ∂Ψ T nd ∂Ψ ∂Ψ , Σnd = ρ F F , B =ρ , Hnd = ρ F , ∂T ∂L ∂β ∂ grada β

∂Φ ∂Φ ∂Φ , Bd = , Hd = , ∂ (D(U)) ∂ (d β /dt) ∂ (gradx d β /dt) ∂Φ , ∂ (gradx T )

(B.8)

290

B The Small Perturbation Assumption

and Σ = Σnd + Σd , B = Bnd + Bd , H = Hnd + Hd . We assume the reference configuration is stress free

∂Ψ (0, β , grada β ) = 0. ∂L This assumption makes possible that Σ nd is of order η . In the examples, it is to be checked if the internal forces are of order η or η 2 as required. The small perturbation interior forces may be introduced with small perturbation ˆ βˆ , gradβˆ ) and Φˆ (gradθˆ , free energy and pseudopotential of deformation Ψˆ (ε (u), ˆ ˆ ∂ β /∂ t, grad∂ β /∂ t). In this setting, the interior forces satisfy either Assumption 2.1 or 2.1bis, provided the reaction forces, i.e., the elements of the subdifferential sets, satisfy the same assumption. This property does not result from the constitutive law. It is either an assumption or the consequence of a balance equation.

B.7 Example 1: The Stefan Problem We assume that temperature θ = T − T0 and β satisfy Assumption 2.1. The free energy and pseudopotential of dissipation are L Ψ (T, βˆ ) = −CT ln T − βˆ (T − T0 ) + I(βˆ ), T0

Φ (grad T, T ) =

λ (grad T )2 . 2T

The constitutive laws are B∈−

L (T − T0 ) + ∂ I(βˆ ), T0

s = C(1 + lnT ) = C(1 + ln T0 ) + Q=−

C C (T − T0 ) = s(T0 ) + (T − T0 ), T0 T0

λ λ grad T = − grad θ . T T0 + θ

They give B∈−

L θ + ∂ I(β ), T0

s − s(T0 ) = sˆ − s(T0 ) = ˆ = − λ grad θˆ . Q T0

C ˆ θ, T0

B.8 Example 2: The Damage Predictive Theory

291

The equation of motion for the microscopic motion is −B = 0. It results that the two quantities B = Bˆ and s − s(T0 ) satisfy Assumption 2.1.

B.8 Example 2: The Damage Predictive Theory We assume βˆ is of order 0. The free energy and pseudopotential are C ˆ 2 βˆ ˆ βˆ , grad βˆ , θˆ ) = − ˆ 2 + 2μe ε (u) ˆ : ε (u)} ˆ Ψˆ (ε (u), θ + {λe (trε (u)) 2T0 2 k +w(1 − βˆ ) + I(β ) + (grad βˆ )2 , 2  ˆ ˆ 2 ∂ β λ ∂ β ∂ βˆ c 2 )= ). Φˆ (grad θˆ , (grad θˆ ) + + I−( ∂t 2T0 2 ∂t ∂t They give the constitutive laws ˆ = βˆ (λe (trε (u)1 ˆ + 2μeε (u)) ˆ , Σ 1 ∂ βˆ ∂ βˆ ˆ 2 + 2μe ε (u) ˆ : ε (u)} ˆ −w+c + ∂ I(βˆ ) + ∂ I−( ), Bˆ ∈ {λe (trε (u)) 2 ∂t ∂t ˆ = − λ grad θˆ , ˆ = k grad βˆ , Q H T0 sˆ =

C ˆ θ. T0

We assume that w is of order η 2 , c and k are of order η . We assume also that the exterior works A and a satisfy Assumption 3bis. They are of order η 2 , (in many engineering problems they are 0). The equation of motion (B.6) show that B is of ˆ βˆ and θˆ are order η 2 . Thus the small perturbation equations for u,

∂ θˆ − λ Δ θˆ = T0 R, in Ωa , ∂t ∂ βˆ ∂ βˆ 1 ˆ 2 c − kΔ βˆ + ∂ I(βˆ ) + ∂ I− ( )  w − {λe (trε (u)) ∂t ∂t 2 ˆ : ε (u)} ˆ + A, in Ω a , +2 μe ε (u)

 ˆ div βˆ (λe (trε (u)1 ˆ + 2 μeε (u)) + f = 0, in Ωa ,

C

292

B The Small Perturbation Assumption

∂ βˆ ∂ θˆ = T0 π , k = a, σ N = g, in ∂ Ω a , ∂N ∂N θˆ (x, 0) = θ 0 (x), βˆ (x, 0) = β 0 (x), in Ω a . λ

Let us note that the small perturbation assumption may be applied partially. For instance, in this example, the quantity −CT ln T may be kept in the free energy and not be replaced by −C θˆ 2 /2T0 (see the free energy of Chap. 6). It is also possible to keep in the equations some quantities of order η 2 . For instance in damage theory, the dissipative quantities 1 T



 Σ :ε ˆd

 ∂ uˆ ∂ βˆ ˆ d + B . ∂t ∂t

may be kept assuming linear viscosity, [38]. To keep this quantity implies clever and sophisticated mathematics to prove mathematical coherency of the model. The necessity to escape the small perturbation assumption leads to difficulties but it is a clear challenge.

B.8.1 Some Numbers for the Equation for the Microscopic Motion Let us give some values of the parameters for concrete damage, [120, 175] w = 10−4 MPa, c = 10−2 MPa s, k = 10−4 MPa m2 , λe  μe  104 MPa. Let us recall, the Lam´e parameters λe and μe are of order η 0 = 1. Thus choosing η = 10−4 , we have w = Aw η 2 with Aw = 1, μe c = Ac η with Ac = 10−2 , μe k = Ak η with Ak = 10−4 . μe In this setting, η represents the intensity of the small deformation. The numerical values show that elasticity and cohesion of concrete, defining damage source 1 ˆ 2 + 2μe ε (u) ˆ : ε (u)}, ˆ w − {λe (trε (u)) 2

B.10 Advantages and Disadvantages of the Small Perturbation Assumption

293

with Lam´e parameters and cohesion w, govern the evolution. The microscopic motion viscosity and local interaction are not as important.

B.9 A Practical Rule From the previous results, one may note that the small perturbation equation may be obtained rapidly by dealing with products XY which appear in an equation, where only the terms of order η p are kept, with the following rule: • Assuming X of order x, Y of order y. • If x + y > p, the product is neglected. • If x + y = p, the product is replaced by Xˆ Yˆ .

B.10 Advantages and Disadvantages of the Small Perturbation Assumption The small perturbation assumption is mainly used in solid mechanics. In civil engineering most of the theories use it. The advantages of the small perturbation assumptions are: 1. The equations are written on a fixed and known domain, Ωa , the reference configuration. 2. The mechanical quantities have an immediate and simple physical meaning as those defined on the actual configuration. They may be easily involved in engineering projects. 3. Mathematical results are available giving the opportunity to have efficient numerical methods. 4. The engineering and practical implications of the predictive theories are often useful and instructive. The everyday success of the elasticity theory is so important that it is often forgotten that it is a small perturbation theory. 5. Let us also mention that the theories without the small perturbation assumption are far from being well established and accepted for engineering purposes. A disadvantage is to add a new subjective assumption, which means more difference between nature and the predictive theories. The small perturbation assumption is a new cause of maladjustment of a theory to predict the evolution of structures. A careful and useful way for the engineer to deal with predictive theories is to be convinced that they are only models which give schematic information. In case the information is not good or useful, the theory is used either in an unadapted or in an extreme situation. We think that a theory must not be used in extreme situations. In case an extreme situation occurs, the theory has to be quit and replaced

294

B The Small Perturbation Assumption

by a new theory taking into account the new phenomenon and quantities which are responsible for the extreme situation. In this careful point of view, the constant comparison of theoretical results with experiments is the guide to judge a predictive theory. Another disadvantage of the small perturbation assumption is that the rules are vague and do not have a sound mathematical basis. This is due to the basic idea which is to find approximation of the large perturbation solutions which are not defined because the large perturbation theories, for instance the constitutive laws, are far from being known. It is very daring to approximate something which is not well known. Let us also note that the small perturbation theory is not a linearization process because of the internal constraints which make compulsory to deal with non-smooth equations where the notion of linearization is not pertinent.

References

1. M. Achenbach, 1989, A model for an alloy with shape 2. J. Aguirre-Puente, M. Fr´emond, 1975, Frost propagation in wet porous media, in Applications of Methods of Functional Analysis to Problems in Mechanics, P. Germain, B. Nayroles eds, Springer, Berlin, Heidelberg. 3. J. Aguirre-Puente, M. Fr´emond, 1976, Frost and water propagation in porous media, Second Conference on Soil Water Problems in Cold Regions, Edmonton. 4. O. Allix, P. Ladev´eze, A. Corigliano, 1995, Damage analysis of interlaminar fracture specimens, Compos. Struct., 31, 61–74. 5. L. Ambrosio, N. Fusco, D. Pallara, 2000, Special functions of bounded variations and free discontinuity problems, Oxford University Press, Oxford. 6. H. Amor, J. J. Marigo, C. Maurini, 2009, Regularized formulation of the variational brittle fracture with unilateral contact: numerical experiments, J. Mech. Phys. Solids, 57, 1209–1229. 7. K. T. Andrews, A. Klarbring, M. Shillor, S. Wright, 1997, A dynamic contact problem with friction and wear, Int. J. Eng. Sci., 35, 1291–1309. 8. S. S. Antman, 1995, Nonlinear problems of elasticity, Springer Verlag, New York. 9. A. Anthoine, P. Pegon, 1996, Numerical analysis and modelling of the damage and softening of brick masonry, in Numerical analysis and modelling of composite materials, J. W. Ball ed, Blacke Academic and Professional, London. 10. P. Argoul, K. Benzarti, F. Freddi, M. Fr´emond, Ti Hoa Tam Nguyen, 2011, A damage model to predict the durability of bonded assemblies. Part II: parameter identification and preliminary results for accelerated ageing tests, doi:10.1016/j.conbuildmat.2009.12.014, Constr. and Build. Mater., 25, 2, 556–567. 11. L. Ascione, A. Grimaldi, F. Maceri, 1981, Modeling and analysis of beams on tensionless foundations, Int. J. for Model. Simul. 12. F. Ascione, M. Fr´emond, 2009, Phase change with voids and bubbles, Vietnam J. Mech., VAST, 31, 3, 263–278. 13. M. Aso, M. Fr´emond, N. Kenmochi, 2004, Phase change problems with temperature dependent constraints for velocity of volume fractions, J. of Nonlinear Anal. Ser. A, Theor. and Methods, 60, 1003–1023. 14. M. Aso, M. Fr´emond, N. Kenmochi, 2004, Quasi-variational evolution problems for irreversible phase change, Math. Sci. and Appl., 20, 517–535. 15. F. Ascione, 2009, Mechanical behavior of FRP adhesive joints: A theoretical model, Composites Part B, 40, 116–124. 16. F. Ascione, 2009, Ultimate behaviour of adhesively bonded FRP lap joints, Composites Part B, 40, 107–115.

M. Fr´emond, Phase Change in Mechanics, Lecture Notes of the Unione Matematica Italiana 13, DOI 10.1007/978-3-642-24609-8, © Springer-Verlag Berlin Heidelberg 2012

295

296

References

17. F. Ascione, G. Mancusi, 2010, Axial/bending coupled analysis for FRP adhesive lap joints, Mech. of Adv. Mater. and Struct., 17, 85–98. 18. H. Attouch, G. Buttazzo, G. Michaille, 2004, Variational analysis in Sobolev and BV spaces, Application to PDE and optimization, MPS/SIAM Series in Optimization. 19. F. Auricchio, L. Petrini, 2004, A three-dimensional model describing stress-temperature induced solid phase transformations. Part I: solution algorithm and boundary value problems, Int. J. Num. Methods Eng., 61, 807–836. 20. F. Auricchio, L. Petrini, 2004, A three-dimensional model describing stress-temperature induced solid phase transformations. Part II: thermomechanical coupling and hybrid composite applications, Int. J. Num. Methods Eng., 61, 716–737. 21. H. Bailly, D. Menessier, C. Prunier, R. Dautray, 1996, Le combustible nucl´eaire des r´eacteurs a` eau sous pression et des r´eacteurs a` neutrons rapides : conception et comportement, ´ Collection du Commissariat a` l’Energie Atomique, S´erie Synth`eses, Eyrolles, Paris. 22. X. Balandraud, E. Ernst, E. So´os, 1999, Ph´enom`enes rh´eologiques dans les alliages a` m´emoire de forme, C. R. Acad. Sci, Paris, II, 327, 1, 33–39. 23. J. M. Ball, R. D. James, 1992, Theory for the microstructure of martensite and applications, Proc. Int. Conf. on Martensitic Transformations, C. M. Wayman, J. Perkings eds, Monterey. 24. A. Barbu, G. Martin, 1993, Radiation effects in metals and alloys, Solid State Phenom., 30 and 31, 179–228. 25. G. I. Barenblatt, 1962, The mathematical theory of equilibrium cracks in brittle fracture, Adv. Appl. Mech., 7, 55–129. 26. E. Benvenuti, G. Borino, A. Tralli, 2002, A Thermodynamically consistent nonlocal formulations for damaging materials, Eur. J. of Mech. A/Solids, 21, 535–553. 27. K. Benzarti, M. Fr´emond, P. Argoul, Ti Hoa Tam Nguyen, 2009, Durability of bonded assemblies. A predictive theory coupling bulk and interfacial damage mechanisms, Eur. J. of Environmental and Civil Eng., 13, 9, 1141–1151. 28. K. Benzarti, F. Freddi, M. Fr´emond, 2011, A damage model to predict the durability of bonded assemblies. Part I: debonding behaviour of FRP strengthened concrete structures, doi:10.1016/j.conbuildmat.2009.10.018, Constr. and Build. Mater., 25, 2, 547–555. 29. C. Berriet, C. Lexcellent, B. Raniecki, A. Chrysochoos, 1992, Pseudoelastic behaviour analysis by infrared thermography and resistivity measurements of polycristalline shape memory alloys, ICOMAT 92, Monterey. 30. M. Berveiller, E. Patoor, 1997, Micromechanical modelling of the thermomechanical behaviour of shape memory alloys, Mechanics of solids with phase change, M. Berveiller, F. Fischer eds, Springer Verlag. 31. P. Bisegna, F. Maceri, 1998, Delamination of active layers in piezoelectric laminates, IUTAM Symposium Variations de domaines et fronti`eres libres en m´ecanique des solides, Kluwer, Amsterdam. 32. D. Blanchard, M. Fr´emond, 1984, The Stefan problem. Computing without the free boundary, Int. J. Num. Methods Eng., 20, 757–771. 33. D. Blanchard, M. Fr´emond, 1985, Soils Frost Heaving and Thaw Settlement, Ground Freezing 85, S. Kinosita and M. Fukuda eds, 209–216, Balkema, Rotterdam. 34. C. Bonacina, G. Comini, A. Fasano, M. Primicerio, 1973, Numerical solution of phase-change problems, Int. J. Heat Mass Transfer 16, 1825–1832. 35. E. Bonetti, 2002, Global solution to a nonlinear phase transition model with dissipation, Adv. Math. Sci. Appl. 12, 355–376. 36. E. Bonetti, 2003, Global solvability of a dissipative Fr´emond model for shape memory alloys. Part I: mathematical formulation and uniqueness, Quart. Appl. Math., 61, 759–781. 37. E. Bonetti, 2004, Global solvability of a dissipative Fr´emond model for shape memory alloys. Part II: existence, Quart. Appl. Math., 62, 53–76. 38. E. Bonetti, G. Bonfanti, 2003, Existence and uniqueness of the solution to a 3D thermoviscoelastic system, Electron. J. Diff. Eqn., 50, 1–15. 39. E. Bonetti, G. Bonfanti, 2005, Asymptotic analysis for vanishing acceleration in a thermoviscoelastic system, Quad. Semin. Mat. Brescia, 16, Abstract and Appl. Anal., 105–120.

References

297

40. E. Bonetti, P. Colli, M. Fr´emond, 2003, A phase field model with thermal memory governed by the entropy balance, Math. Models and Methods in Appl. Sci., 13, 231–256. 41. E. Bonetti, M. Fr´emond, 2003, A phase transition model with the entropy balance, Math. Methods in the Appl. Sci., 26, 539–556. 42. E. Bonetti, M. Fr´emond, 2004, Damage theory: microscopic effects of vanishing macroscopic motions, Comput. and Appl. Math., 22, 3, 1–21. 43. E. Bonetti, M. Fr´emond, 2010, Analytical results on a model for damaging in domains and interfaces, doi: 10.1051/cocv/2010033, pubished on line 18 August 2010, Control, Optim. and Calc. of Var. 44. E. Bonetti, M. Fr´emond, Ch. Lexcellent, 2004, Modelling shape memory alloys, J. de Phys. IV France, 115, 383–390. 45. E. Bonetti, M. Fr´emond, Ch. Lexcellent, 2007, Hydrogen storage: modelling and analytical results, Appl. Math. and Optim., 55, 1, 1–59. 46. E. Bonetti, G. Schimperna, 2004, Local existence to Fr´emond’s model for damage in elastic materials, Cont. Mech. Therm., 16, 319–335. 47. G. Bonfanti, M. Fr´emond, F. Luterotti, 2000, Global solution to a nonlinear system for irreversible phase changes, Adv. Math. Sci. Appl. 10, 1–24. 48. G. Bonfanti, M. Fr´emond, F. Luterotti, 2001, Local solutions to the full model of phase transitions with dissipation, Adv. Math. Sci. Appl. 11, 791–810. 49. G. Bonfanti, M. Fr´emond, F. Luterotti, 2004, Existence and uniqueness results to a phase transition model based on microscopic acceleration and movements, Nonlinear Anal.: Real World Appl., 5, 1, 123–140. 50. G. Borino, B. Failla, F. Parrinello, 2003, A Symmetric nonlocal damage theory, Int. J. Solids and Struct., 40, 3621–3645. 51. A. Bossavit, 1976, D´efinition et calcul d’une perm´eabilit´e e´ quivalente pour l’acier satur´e. Un ´ probl`eme de Stefan a` petit param`etre, Bulletin de la Direction des Etudes et Recherches, S´erie C, 2, 45–58. 52. G. Bouchitt´e, A. Mielke, T. Roub´ıcˇ ek, 2007 A complete damage problem at small strains, Z. Angew. Math. Phys. 60, 2009, 205–236 (published eletronically Nov. 2007). 53. B. Bourdin, G. A. Francfort, J. J. Marigo, 2000, Numerical experiments in revisited brittle fracture, J. Mech. Phys. Solids, 48, 797–826. 54. A. Braides, 1998, Approximation of free-discontinuity problems, Springer-Verlag, Berlin. 55. C. M. Brauner, B. Nicolaenco, M Fr´emond, 1986, Homographic approximations of free boundary problems characterized by variational inequalities. Sci. Comput., Adv. Math. Suppl. Stud., 10, 119–151. 56. M. Brokate, J. Sprekels, 1996, Hysteresis and phase transitions, Appl. Math. Sci. 121, Springer, New York. 57. M. D. Bui, 1978, M´ecanique de la rupture fragile, Masson, Paris. 58. L. Cangemi, 1997, Frottement et adh´erence : mod`ele, traitement num´erique et application a` l’interface fibre/matrice, Th`ese de l’Universit´e d’Aix-Marseille. 59. P. Carrara, D. Ferretti, F. Freddi, G. Rosati, 2011, Shear tests of carbon fiber plates bonded to concrete with control of snap-back, Eng. Fract. Mech., 78, 2663–2678. 60. P. Casal, 1972, La th´eorie du second gradient et la capillarit´e, C. R. Acad. Sci. Paris, A, 274, II, 1571–1574. 61. A. M. Caucci, M. Fr´emond, 2005, Collisions and phase change, Free boundary problems: theory and application, Coimbra. 62. A. M. Caucci, M. Fr´emond, 2009, Thermal effects of collisions: does rain turn into ice when it falls on a frozen ground? J. of Mech. of Mater. and Struct., 4, 2, 225–244, http://pjm.math.berkeley.edu/jomms/2009/4-2/index.xhtml 63. N. Chemetov, 1998, Uniqueness results for the full Fr´emond model of shape memory alloys, Z. Anal. Anwend., 17, 4, 877–892. 64. A. Chrysochoos, H. Pham, O. Maisonneuve, 1993, Une analyse exp´erimentale du comportement d’un alliage a` m´emoire de forme de type Cu-Zn-Al, C. R. Acad. Sci., Paris, 316, II, 1031–1036.

298

References

65. A. Chrysochoos, M. L¨obel, O. Maisonneuve, 1994, Couplages thermom´ecaniques du comportement pseudo´elastique d’alliages Cu-Zn-Al et Ni-Ti, C. R. Acad. Sci., Paris, 320, IIb, 217–223. 66. P. G. Ciarlet, 1988, Mathematical Elasticity Volume I: Three-Dimensional Elasticity, NorthHolland, Amsterdam. 67. M. Cocu, L. Cangemi, M. Raous, 1996, Adhesion and friction model for the fiber/matrix interface of a composite, PD-Vol 76, ESDA Proceedings of the 1996 Engineering Systems Design and Analysis Conference, A. Lagarde, M. Raous eds, Book n◦ G000994-1996, ESDA. 68. P. Colli, M. Fr´emond, O. Klein, 2001, Global existence of a solution to a phase field model for supercooling, Nonlinear Anal.: Real World Appl., 2, 523–539. 69. P. Colli, M. Fr´emond, A. Visintin, 1990, Thermomechanical evolution of shape memory alloys, Q. Appli. Math., XLVIII, 1, 31–47. 70. P. Colli, G. Gentili, C. Giorgi, 1999, Non linear systems describing phase transition models compatible with thermodynamics, Math. Models and Methods in Appl. Sci. 9, 1015–1037. 71. P. Colli, G. Gilardi, M. Grasselli, 1997, Well-posedness of the weak formulation for the phasefield model with memory, Adv. Diff. Equ. 2, 3, 487–508. 72. P. Colli, N. Kenmochi, M. Kubo, 2001, A phase-field model with temperature dependent constraint, J. Math. Anal. Appl. 256, 668–685. 73. P. Colli, F. Luterotti, G. Schimperna, U. Stefanelli, 2002, Global existence for a class of generalized systems for irreversible phase changes, Nonlinear Diff. Equ. Appl., 9, 255–276. 74. P. Colli, J. Sprekels, 1995, Global solution to the full one-dimensional Fr´emond model for shape memory alloys, Math. Methods. Appl. Sci., 18, 371–385. 75. C. Comi, 2001, A non local model with tension and compression damage mechanics, Eur. J. Mech. A/Solids, 20, 1–22. 76. H. da Costa Mattos, M. Fr´emond, E. N. Mamiya, 1992, A simple model of the mechanical behaviour of ceramic like material, Int. J. Solids Struct., 29, 24, 3185–3200. 77. G. Dal Maso, G. Lazzaroni, 2010, Quasistatic crack growth in finite elasticity with noninterpenetration, Ann. Inst. H. Poincar´e, Anal. Non Lin´eaire, 27, 1, 257–290. 78. A. Damlamian, 1977, Some results on the multi-phase Stefan problem, Comm. Partial Diff. Equ. 10, 1017–1044. 79. E. Darque-Ceretti, E. Felder, 2003, Adh´esion et adh´erence, Sciences et thechniques de l’ing´enieur, CNRS Editions, Paris. 80. R. de Borst, J. Mazars, G. Pijaudier-Cabot, J. G. M. van Mier 2001, Durability, time dependence and coupled problems, in Fracture Mechanics of Concrete Structures, R. de Borst, J. Mazars, G. Pijaudier-Cabot, J. G. M. van Mier eds, Balkema, 133–278. 81. G. Del Piero, G. Lancioni, R. March, 2007, A Variational Model for Fracture Mechanics. Numerical Experiments, J. Mech. Phys. Solids, 55, 2513–2537. 82. G. Del Piero, M. Raous, 2010, A unified model for adhesive interfaces with damage, viscosity, and friction, Eur. J. of Mech. A/Solids, 29, 4, 496–507. 83. E. Dimnet, Collisions de solides d´eformables, Th`ese de l’Ecole nationale des Ponts et Chauss´ees, Paris, 2000. 84. E. Dimnet, Collision d’un rocher et d’un arbre, to appear. 85. E. Dimnet, R. Gormaz, 2005, Le ricochet d’une pierre sur l’eau d’un lac, to appear. 86. E. Dimnet, 2000, M. Fr´emond, Instantaneous collisions of solids, ECCOMAS, Barcelone. 87. A. Dupas, N. L. Tran, 1995, Mesure de la teneur en eau non gel´ee dans les milieux poreux par m´ethodes di´electriques, Les sols gel´es : processus thermophysiques et g´eomorphologiques. G´eomorphologie et transferts de surface, Caen. 88. A. Dupas, N. L. Tran, 1996, L’eau dans le g´enie civil, Journ´ees th´ematiques de mesures de l’eau dans les mat´eriaux, CEA-LETI, Grenoble. 89. Ch. Duquennoi, 1992, Etude sur la performance des fondations superficielles isol´ees soumises a` l’action du gel saisonnier des sols, Th`ese de l’Ecole nationale des Ponts et Chauss´ees, Paris. 90. I. Ekeland, R. Temam, 1974, Analyse convexe et probl`emes variationnels, Dunod-GauthierVillars.

References

299

91. M. Fabrizio, C. Giorgi, A. Morro, 2006, A thermodynamic approach to non-isothermal phasefield evolution in continuum physics, Physica D-Nonlinear Phenom., 214, 144–156. 92. M. Fabrizio, 2006, Ginzburg-Landau equations and first and second order phase transitions, Int. J. of Eng. Sci., 44, 529–539. 93. M. Fabrizio, C. Giorgi, M. G. Naso, 2004, Viscoelastic solids of exponential type. I. Minimal representations and controllability, Meccanica, 39, 6, 531–546. 94. M. Fabrizio, C. Giorgi, M. G. Naso, 2004, Viscoelastic solids of exponential type. II. Free energies, stability and attractors, Meccanica, 39, 6, 547–561. 95. F. Falk, 1982, Landau theory and martensitic phase transition, J. de Phys.; Colloque C4, Suppl´ement au n◦ 12, 43. 96. A. Fasano, M. Primicerio, 1988, A phase change model with a zone of coexistence of phases, IMA J. Appl. Math., 4, 31–46. 97. S. Faure, J. M. Ghidaglia, 2011, Violent Flows in Aqueous Foam I: Phys. and Numer. Models, Eur. J. Mech. B/Fluids, 30, 341–359. 98. G. Fichera, 1972, Boundary value problems in elasticity with unilateral constraints, Handbuch der Physik, Vol. VIa/2, Springer Verlag, Berlin. 99. F. Freddi, M. Fr´emond, 2006, Damage in domains and interfaces: a coupled predictive theory, J. Mech. Mater. Struct., 1, 7, 1205–1234. 100. F. Freddi, M. Fr´emond, 2010, Collisions and fractures: a predictive theory, Eur. J. of Mech. A/Solids, doi:10.1016/j.euromechsol.2010.07.006, 29, 998–1007. 101. F. Freddi, G. Royer-Carfagni, 2010, Regularized Variational Theories of Fracture: a Unified Approach, J. Mech. Phys. Solids, 58, 1154–1174. 102. F. Freddi, G. Royer-Carfagni, 2011, Variational fracture mechanics for compressive splitting of masonry-like materials, Ann. Solid and Struct. Mech., 2, 57–67. 103. F. Freddi, M. Savoia, 2008, Analysis of FRP-concrete delamination via boundary integral equations, Eng. Fract. Mech., 75, 6, 1666–1683. 104. M. Fr´emond, 1982, Adh´erence des solides, C.R. Acad. Sci. Paris, II, 295, 769–772. and Equilibre de structures qui adh´erent a` leur support, C.R. Acad. Sci. Paris, II, 295, 913–915. 105. M. Fr´emond, 1982, M´ethodes variationnelles en calcul des structures, Ecole nationale des Ponts et Chauss´ees, Paris. 106. M. Fr´emond, 1982, Adh´esion et contact unilat´eral, in Contact mechanics and wear of rail/wheel Systems, University of Waterloo Press, Waterloo, Canada. 107. M. Fr´emond, 1985, Contact unilateral avec adh´erence. Une th´eorie du premier gradient, in G. Del Piero and F. Maceri eds, Unilateral problems in structural analysis, Springer Verlag. 108. M. Fr´emond, 1987, Adh´erence des solides, J. Mec. Theor. Appl., 6, 3, 383–407. 109. M. Fr´emond, 1987, Mat´eriaux a` m´emoire de forme, C. R. Acad. Sci., Paris, 304, II, 7, 239–244. 110. M. Fr´emond, 1998, L’´education des mat´eriaux a` m´emoire de forme, Rev. Eur. des e´ l´ements finis, 7, 8, 35–46. 111. M. Fr´emond, 2000, Phase change with temperature discontinuities, Gakuto Inter, Ser. Math. Sci. Appl. 14, 125–134. 112. M. Fr´emond, 2002, Non-smooth thermomechanics, Springer-Verlag Berlin. 113. M. Fr´emond, 2002, Damage theory. A macroscopic motion vanishes but its effects remain, Comput. and Appl. Math., 21, 2, 1–14. 114. M. Fr´emond, 2007, Collisions, Edizioni del Dipartimento di Ingegneria Civile, Universit`a di Roma “Tor Vergata”, ISBN 978-88-6296-000-7. 115. M. Fr´emond, R. Gormaz, J. San Martin, 2003, Collision of a solid with an uncompressible fluid, Theoretical and Comput. Fluid Dyn., 16, 405–420. 116. M. Fr´emond, N. Kenmochi, 2006, Damage of a viscous locking material, Adv. in Math. Sci. and Appl., 16, 2, 697–716. 117. M. Fr´emond, K. L. Kuttler, B. Nedjar, M. Shillor, 1998, One dimensional models of damage, Adv. Math. Sci. Appl., 8, 2, 541–570. 118. M. Fr´emond, K. L. Kuttler, M. Shillor, 1999, Existence and uniqueness of solutions for a dynamic one-dimensional damage model, J. Math. Anal. Appl., 229, 271–294.

300

References

119. M. Fr´emond, S. Myasaki, 1996, Shape memory alloys, CISM, Courses and Lectures n◦ 351, Springer Verlag, Wien. 120. M. Fr´emond, B. Nedjar, 1993, Endommagement et principe des puissances virtuelles, C. R. Acad. Sci. Paris, II, 317, 7, 857–864. 121. M. Fr´emond, B. Nedjar, 1995, Damage in concrete: the unilateral phenomenon, Nuclear Eng. and Design 156, Elsevier Science, 323–335. 122. M. Fr´emond, B. Nedjar, 1996, Damage, gradient of damage and principle of virtual power, Int. J. Solids Struct., 33, 8, 1083–1103. 123. M. Fr´emond, B. Nedjar, 1998, Damage in fast dynamics, in Civil infrastructures systems: intelligent renewal, F. Casciati, F. Maceri, M.P. Singh, P. Spanos eds. World Scientific, Singapore. 124. M. Fr´emond, E. Rocca, 2006, Well-posedness of a phase transition model with the possibility of voids, Math. Models and Methods in Appl. Sci., 4, 16, 559–586. 125. M. Fr´emond, E. Rocca, 2008, Solid liquid phase changes with different densities, Q. Appl. Math., 66, 609–632. 126. M. Fr´emond, E. Rocca, 2009, A model for shape memory alloys with the possibility of voids, Discret. and Continuous Dyn. Syst., Ser. A, 27, 4, http://aimsciences.org/ journals/displayPapers1.jsp?pubID=357. 127. M. Fr´emond, A. Visintin, 1985, Dissipation dans le changement de phase. Surfusion. Changement de phase irr´eversible, C. R. Acad. Sci. Paris, II, 301, 1265–1268. 128. Gel des sols et des chauss´ees, 1979, M. Fr´emond, P. J. Williams eds, Ecole nationale des Ponts et Chauss´ees, Paris. 129. A. Gerard, S. Baste, 1991, Nonlinear behavior of ceramic matrix composites during loadunload cycles, C. R. Acad. Sci. Paris, 313, II, 1361–1466. 130. P. Germain, Q. S. Nguyen, P. Suquet, 1983, Continuum Thermodynamics, J. Appl. Mech., ASME 50, 1010–1021. 131. P. Germain, 1973, La m´ethode des puissances virtuelles en m´ecanique des milieux continus, J. Mec., 12, 2, 235–274. 132. P. Germain, 1973, M´ecanique des milieux continus, Masson, Paris. 133. H. Ghidouche, 1990, Etude math´ematique de mod`eles de changement de phase dans les milieux dissipatifs et dans un milieu poreux satur´e, Th`ese de Doctorat d’Etat, Universit´e de Paris Nord, Villetaneuse. 134. D. Gonzalez, K. Benzarti, L. Gonon, L. De Baynast, 2005, Durability of the concrete epoxy/ adhesive bond : micro- and macro-scale investigations, 3rd International Conference Composites in Construction. 135. Ground Freezing 94, 1994, M. Fr´emond ed., Balkema, Rotterdam. 136. G. Gu´enin, 1986, Alliages a` m´emoire de forme, Techniques de l’Ing´enieur, M 530, Paris. 137. M. E. Gurtin, 1996, Generalized Ginzburg-Landau and Cahn-Hilliard equations based on a microforce balance, Phys. D, 92, 178–192. 138. D. Halm, A. Dragon, 1997, An anisotropic model of damage and frictional sliding for brittle materials, Eur. J. of Mech. A/Solids, 17, 3, 439–460. 139. B. Halphen, Nguyen Quoc Son, 1975, Sur les mat´eriaux standards g´en´eralis´es, J. Mec., 14, 1, 39–63. 140. G. J. Haltiner, Numerical Weather Prediction, 1971, J. Wiley, New York. 141. P. Humbert, G. Fezans, A. Dubouchet, D. Remaud, 2005, CESAR-LCPC, un progiciel de calcul d´edi´e au g´enie civil, Bulletin des Laboratoires des Ponts et Chauss´ees, 256–257. 142. V. Huon, B. Cousin, O. Maisonneuve, 2003, Study of thermal and kinematic phenomena associated with quasi-static deformation and damage process of some concretes, dans Novel approaches in civil engineering, M. Fr´emond, F. Maceri eds. Springer-Verlag, Berlin, 187–201. 143. M. Ishii, T. Hibiki, 2006, Thermo-Fluid Dynamics of Two-Phase Flow, Springer, New-York. 144. J. W. Jerome, 1983, Approximation of nonlinear evolution systems, Ser. Math. in Sci. and Eng. 164, Academic Press.

References

301

145. L.M. Kachanov, 1958, On the life-time under creep conditions. Izv. Akad. Nauk SSSR 8, 26–31 (in Russian). 146. N. Kikuchi, Y. Ichikama, 1979, Numerical methods for a two-phase Stefan problem by variational inequalities, Int. J. Num. Methods. Eng. 14, 1221–1239. 147. R. Kling, 1967, Thermodynamique g´en´erale et applications, Editions Technip, Paris. 148. P. Ladev`eze, 1983, Sur une th´eorie de l’endommagement anisotrope. Internal report n◦ 34, Laboratoire de M´ecanique et Technologie, Ecole normale sup´erieure de Cachan, France. 149. G. Lam´e, B. P. Clapeyron, 1831, M´emoire sur la solidification par refroidissement d’un globe liquide, Ann. de Chim. et de Phys., 47, 250–256. 150. P. Lebret, A. Dupas, M. Clet, J. P. Coutard, J. P. Lautridou, S. Courbouleix, M. Garcin, M. L´evy, B. Van Vliet Lano¨e, 1994, Modelling of permafrost thickness during the late glacial stage in France: preliminary results, Can. J. Earth. Sci., 31, 956–968. 151. J. Lemaˆıtre, J. L. Chaboche, 1988, M´ecanique des mat´eriaux solides, Dunod, Paris. 152. J. Lemaˆıtre, 1992, A course on damage mechanics, Springer Verlag, Berlin. 153. R. Lassoued, 2000, Comportement hivernal des chauss´ees. Mod´elisation thermique, Th`ese de l’Ecole nationale des Ponts et Chauss´ees, Paris. 154. M. Latroche, 2004, Structural and thermodynamic properties of metal hydrides used for energy storage, J. Phys. Chem. Solids, 65, 517–522. 155. S. Leclercq, 1995, De la mod´elisation thermom´ecanique et de l’utilisation des alliages a` m´emoire de forme, Th`ese de l’Universit´e de Franche-Comt´e, Besanc¸on. 156. S. Leclercq, C. Lexcellent, 1996, General macroscopic description of the thermomechanical behaviour of shape memory alloys, J. Mech. Phys. Solids., 44, 953–981. 157. C. Lexcellent, C. Licht, 1991, Some remarks on the modelling of the thermomechanical behaviour of shape memory alloys, J. de Phys., Colloque C4, 1, 35–39. 158. J. L. Lions, G. Stampacchia, 1967, Variational inequalities, Comm. Pure Appli. Math., 20, 493–519. 159. M. L¨obel, 1994, Caract´erisation thermom´ecanique d’alliages a` m´emoire de forme de type NiTi et CuZnAl. Domaine de transition et cin´etique de changement de phase, Th`ese de l’Universit´e des Sciences et des Techniques du Languedoc, Montpellier. 160. F. Luterotti, G. Schimperna, U. Stefanelli, 2001, Existence result for a nonlinear model related to irreversible phase changes, Math. Models and Methods in Appl. Sci. 11, 1–17. 161. F. Maceri, R. Toscano, A. Maceri, 1979, Alcuni problemi di vincolo unilaterale per sistemi di travi linearmente elastici, nota I e nota II, Rend. Acc. Naz. dei Lincei, Roma. 162. J. P. Magnan, 1997, M´ecanique des sols et des roches, Ecole nationale des Ponts et Chauss´ees, Paris. 163. F. Maceri, M. Marino, G. Vairo, 2012, Integrated mechanical models for collagenous biostructures at different length scales, Proceedings of MATHMOD 2012 - 7th Vienna International Conference on Mathematical Modelling. 164. S. Marfia, E. Sacco, J. N. Reddy, 2003, Superelastic and shape memory effects in laminated shape-memory-alloy beams, AIAA, 41, 1, 100–109. 165. G. A. Maugin, 1990, Internal variables and dissipative structures. J. Non-Equilib. Thermodyn., 15, 2, 173–192. 166. D. Maugis, 1999, Contact, adherence and rupture of elastic solids, Springer Verlag, Berlin, Heidelberg. 167. D. Maugis, M. Barquins, 1978, Fracture mechanics and the adherence of viscoelastic bodies, J. Phys. D., Appl. Phys., 11, 1989–2033. 168. Mechanical Modelling and Computational Issues in Civil Engineering Approaches in civil engineering, 2005, M. Fr´emond, F. Maceri eds, Springer, Berlin. 169. A. Mielke, T. Roub´ıcˇ ek, 2003, A rate-independent model for inelastic behavior of shapememory alloys, Multiscale Model. and Simul. 1, 571–597. 170. Mod´elisation des alliages a` m´emoire de forme, 1998, F. Lebon coordonnateur, Rev. Eur. des e´ l´ements finis, 7, 8. 171. J. J. Moreau, 1970, Sur les lois de frottement, de viscosit´e et de plasticit´e, C. R. Acad. Sci., Paris, 271, 608–611.

302

References

172. J. J. Moreau, 2003, Fonctionnelles convexes, Edizioni del Dipartimento di Ingegneria Civile, Universit`a di Roma “Tor Vergata”, 2003, ISBN 978-88-6296-001-4 and S´eminaire sur les e´ quations aux d´eriv´ees partielles, Coll`ege de France, Paris, 1966. 173. I. M¨uller, 1985, Pseudo elasticity in shape memory alloys. An extreme case of thermoelasticity. Proc. Thermoelasticita Finita, Acc. Naz. dei Lincei. 174. B. Nayroles, 1973, Point de vue alg´ebrique. Convexit´e et int´egrandes convexes en m´ecanique des solides, C.I.M.E., Bressanone, Ed. Cremonese. 175. B. Nedjar, 1995, M´ecanique de l’endommagement. Th´eorie du premier gradient et application au b´eton, Th`ese de l’Ecole nationale des Ponts et Chauss´ees, Paris. 176. M. Negri, 2010, A comparative analysis on variational models for quasi-static brittle crack propagation. Adv. Calc. Var. 3, 149–212. 177. Nguyen Quoc Son, 2010, On standard dissipative gradient models, Annals of Solid and Structural Mech., doi: 10.1007/s12356-010-0006-0 1, 2, 79–86. 178. Nguyen Quoc Son, Z. Moumni, 1995, Sur une mod´elisation du changement de phases solides, C. R. Acad. Sci., Paris, II, 321, 3, 87–92. 179. M. Niezgodka, J. Sprekels, 1991, Convergent numerical approximation of the thermomechanical phase transitions in shape memory alloys, Numer. Math. 58, 759–778. 180. Novel approaches in civil engineering, 2003, M. Fr´emond, F. Maceri eds, Springer-Verlag, Berlin. 181. P. D. Panagiotopoulos, 1985, Inequality problems in mechanics and applications, Birkha¨user Verlag, Basel. 182. L. Paoli, 1993, Analyse num´erique de vibrations avec contraintes unilat´erales, Th`ese de l’Universit´e Claude Bernard, Lyon. 183. E. Patoor, M. Berveiller, 1990, Les alliages a` m´emoire de forme, Herm`es, Paris. 184. N. Pede, P. Podio-Guidugli, G. Tomassetti, 2006, Balancing the force that drives the peeling of an adhesive tape, Il Nuovo Cimento, 121 B, 5, 531–543. 185. H. Pham, 1994, Analyse thermom´ecanique d’un alliage a` m´emoire de forme de type Cu-ZnAl, Th`ese de l’Universit´e des Sciences et des Techniques du Languedoc, Montpellier. 186. P. Podio-Guidugli, 2005, Peeling tapes, in Mechanics of Material Forces, P. Steinmann and G.A. Maugin eds, vol. 11 of Advances in Mech. and Math., Springer, 253–260. 187. N. Point, 1988, Unilateral contact with adhesion, Math. Methods Appl. Sci., 10, 367–381. 188. H. Pron, 2000, Application des effets photothermiques et thermom´ecaniques a` l’analyse des contraintes appliqu´ees et r´esiduelles, th`ese de l’Universit´e de Reims Champagne-Ardenne. 189. H. Pron, J. F. Henry, R. Bouferra, C. Bissieux, J. L. Beaudouin, 2002, Etude par thermographie infrarouge du grenaillage de pr´econtrainte ou shot-penning, Congr`es franc¸ais de thermique, SFT 2002, Vittel. 190. K. R. Rajagopal, T. Roub´ıcˇ ek, 2003, On the effect of dissipation in shape-memory alloys. Nonlinear Anal.: Real World Appl., 4, 581–597 191. B. Raniecki, C. Lexcellent, K. Tanaka, 1992, Thermodynamics models of pseudoelastic behaviour of shape memory alloys, Arch. Mech., 44, 3, 261–284. 192. A. Rigolot, 1976, Sur une th´eorie asymptotique des poutres droites, Th`ese de Doctorat d’Etat, Universit´e Pierre et Marie Curie, Paris. 193. A. Rigolot, 1977, D´eplacements finis et petites d´eformations des poutres droites : Analyse asymptotique de la solution a` grande distance des bases, J. Mec. Appl., 1, 2. 194. J. F. Rodrigues, Obstacle problems in mathematical physics, North Holland, 1987. 195. P. P. Rodr´ıguez, A. Ibarra, A. Iza-Mendia, V. Recarte, J. I. P´erez-Landaz´abal, J. San Juan and M. L. N´o, 2004, Influence of thermomechanical processing on the microstructure of Cu-based shape memory alloys produced by powder metallurgy, Mater. Sci. and Eng. A, 378, 263–268. 196. K. Saanouni, J. L. Chaboche, P. M. Lesne, 1988, Creep crack growth prediction by a non local damage formulation, in Cracking and damage, strain localisation and size effects, J. Mazars, Z. P. Bazant eds, Elsevier, Amsterdam, 404–414. 197. J. Salenc¸on, 1988, M´ecanique des milieux continus, Ellipses, Paris. 198. J. San Martin, 1996, Gel des sols en pr´esence de solut´es, Th`ese de l’Universit´e Pierre et Marie Curie, Paris.

References

303

199. A. Signorini, 1925, Un teorema die esistenza ed unicita, nello studio dei materiali poco resistenti a trazione, Rend. Accad. Naz. Lincei 2, 401–406. ¨ 200. J. Stefan, 1890, Uber die Theorie der Eisbildung, ins besonders uˆ ber de Eisbildung in Polarmeere, Sitzungsberichte der Mathematish-Naturwissenschaftlichen Klasse der Kaisserlichen Akademie der Wissenschaft, Vol. XCVIII, n◦ IIa, Wien. 201. D. Stoianovici, Y. Hurmuzlu, 1996, A critical study of the applicability of rigid-body collision theory, ASME, J. Appl. Mech., 63, 307–316. 202. Technologie des alliages a` m´emoire de forme, 1994, M. Berveiller, E. Patoor eds, Herm`es, Paris. 203. M. P. Thaveau, 2005, Flexion quatre points de prismes en b´eton coll´es, LCPC internal report. 204. J. M. Truong Dinh Tien, 1990, Contact avec adh´erence, Th`ese de l’Universit´e Pierre et Marie Curie, Paris. 205. G. Tomassetti, 2011, Quasistatic damage evolution in thin plate-like bodies, in preparation. 206. N. Valoroso, L. Champaney, 2006, A damage-mechanics-based approach for modelling decohesion in adhesively bonded assemblies, Eng. Fract. Mech., doi:10.1016/j.engfracmech.2006. 04.029, 73, 18, 2774–2801. 207. A. Visintin, 1996, Models of phase transition, Birk¨auser. 208. A. Visintin, 2001, A new approach to evolution, C.R. Acad. Sci. Paris, I, 332, 233–238. 209. J. M. Wallace, P. V. Hobbs, 2006, Atmospheric science, second edition, Elsevier. 210. G. W¨orsching, 1995, Numerical simulation of the Fr´emond model for shape memory alloys, Gatuko Int. series: Math. Sci. and Appl., Vol. 7, Non-linear analysis and Appl., N. Kenmochi, M. Niesgodka, P. Strzelecki eds, Tokyo, 425–433.

Editor in Chief: Franco Brezzi Editorial Policy 1. The UMI Lecture Notes aim to report new developments in all areas of mathematics and their applications - quickly, informally and at a high level. Mathematical texts analysing new developments in modelling and numerical simulation are also welcome. 2. Manuscripts should be submitted (preferably in duplicate) to Redazione Lecture Notes U.M.I. Dipartimento di Matematica Piazza Porta S. Donato 5 I – 40126 Bologna and possibly to one of the editors of the Board informing, in this case, the Redazione about the submission. In general, manuscripts will be sent out to external referees for evaluation. If a decision cannot yet be reached on the basis of the first 2 reports, further referees may be contacted. The author will be informed of this. A final decision to publish can be made only on the basis of the complete manuscript, however a refereeing process leading to a preliminary decision can be based on a pre-final or incomplete manuscript. The strict minimum amount of material that will be considered should include a detailed outline describing the planned contents of each chapter, a bibliography and several sample chapters. 3. Manuscripts should in general be submitted in English. Final manuscripts should contain at least 100 pages of mathematical text and should always include – a table of contents; – an informative introduction, with adequate motivation and perhaps some historical remarks: it should be accessible to a reader not intimately familiar with the topic treated; – a subject index: as a rule this is genuinely helpful for the reader. 4. For evaluation purposes, manuscripts may be submitted in print or electronic form (print form is still preferred by most referees), in the latter case preferably as pdf- or zipped ps-files. Authors are asked, if their manuscript is accepted for publication, to use the LaTeX2e style files available from Springer’s web-server at ftp://ftp.springer.de/pub/tex/latex/svmonot1/ for monographs and at ftp://ftp.springer.de/pub/tex/latex/svmultt1/ for multi-authored volumes 5. Authors receive a total of 50 free copies of their volume, but no royalties. They are entitled to a discount of 33.3% on the price of Springer books purchased for their personal use, if ordering directly from Springer. 6. Commitment to publish is made by letter of intent rather than by signing a formal contract. Springer-Verlag secures the copyright for each volume. Authors are free to reuse material contained in their LNM volumes in later publications: A brief written (or e-mail) request for formal permission is sufficient.